DeepSeeded: Volumetric Segmentation of Dense Cell Populations with a Cascade of Deep Neural Networks in Bacterial Biofilm Applications.

Accurate and automatic segmentation of individual cell instances in microscopy images is a vital step for quantifying the cellular attributes, which can subsequently lead to new discoveries in biomedical research. In recent years, data-driven deep learning techniques have shown promising results in this task. Despite the success of these techniques, many fail to accurately segment cells in microscopy images with high cell density and low signal-to-noise ratio. In this paper, we propose a novel 3D cell segmentation approach DeepSeeded, a cascaded deep learning architecture that estimates seeds for a classical seeded watershed segmentation. The cascaded architecture enhances the cell interior and border information using Euclidean distance transforms and detects the cell seeds by performing voxel-wise classification. The data-driven seed estimation process proposed here allows segmenting touching cell instances from a dense, intensity-inhomogeneous microscopy image volume. We demonstrate the performance of the proposed method in segmenting 3D microscopy images of a particularly dense cell population called bacterial biofilms. Experimental results on synthetic and two real biofilm datasets suggest that the proposed method leads to superior segmentation results when compared to state-of-the-art deep learning methods and a classical method.

Targeting Iron - Respiratory Reciprocity Promotes Bacterial Death.

Discovering new bacterial signaling pathways offers unique antibiotic strategies. Here, through an unbiased resistance screen of 3,884 gene knockout strains, we uncovered a previously unknown non-lytic bactericidal mechanism that sequentially couples three transporters and downstream transcription to lethally suppress respiration of the highly virulent P. aeruginosa strain PA14 - one of three species on the WHO's 'Priority 1: Critical' list. By targeting outer membrane YaiW, cationic lacritin peptide 'N-104' translocates into the periplasm where it ligates outer loops 4 and 2 of the inner membrane transporters FeoB and PotH, respectively, to suppress both ferrous iron and polyamine uptake. This broadly shuts down transcription of many biofilm-associated genes, including ferrous iron-dependent TauD and ExbB1. The mechanism is innate to the surface of the eye and is enhanced by synergistic coupling with thrombin peptide GKY20. This is the first example of an inhibitor of multiple bacterial transporters.

Single-molecule tracking to determine the abundances and stoichiometries of freely-diffusing protein complexes in living cells: Past applications and future prospects.

Most biological processes in living cells rely on interactions between proteins. Live-cell compatible approaches that can quantify to what extent a given protein participates in homo- and hetero-oligomeric complexes of different size and subunit composition are therefore critical to advance our understanding of how cellular physiology is governed by these molecular interactions. Biomolecular complex formation changes the diffusion coefficient of constituent proteins, and these changes can be measured using fluorescence microscopy-based approaches, such as single-molecule tracking, fluorescence correlation spectroscopy, and fluorescence recovery after photobleaching. In this review, we focus on the use of single-molecule tracking to identify, resolve, and quantify the presence of freely-diffusing proteins and protein complexes in living cells. We compare and contrast different data analysis methods that are currently employed in the field and discuss experimental designs that can aid the interpretation of the obtained results. Comparisons of diffusion rates for different proteins and protein complexes in intracellular aqueous environments reported in the recent literature reveal a clear and systematic deviation from the Stokes-Einstein diffusion theory. While a complete and quantitative theoretical explanation of why such deviations manifest is missing, the available data suggest the possibility of weighing freely-diffusing proteins and protein complexes in living cells by measuring their diffusion coefficients. Mapping individual diffusive states to protein complexes of defined molecular weight, subunit stoichiometry, and structure promises to provide key new insights into how protein-protein interactions regulate protein conformational, translational, and rotational dynamics, and ultimately protein function.

Dimerization of iLID optogenetic proteins observed using 3D single-molecule tracking in live E. coli.

3D single-molecule tracking microscopy has enabled measurements of protein diffusion in living cells, offering information about protein dynamics and cellular environments. For example, different diffusive states can be resolved and assigned to protein complexes of different size and composition. However, substantial statistical power and biological validation, often through genetic deletion of binding partners, are required to support diffusive state assignments. When investigating cellular processes, real-time perturbations to protein spatial distributions is preferable to permanent genetic deletion of an essential protein. For example, optogenetic dimerization systems can be used to manipulate protein spatial distributions that could offer a means to deplete specific diffusive states observed in single-molecule tracking experiments. Here, we evaluate the performance of the iLID optogenetic system in living E. coli cells using diffraction-limited microscopy and 3D single-molecule tracking. We observed a robust optogenetic response in protein spatial distributions after 488 nm laser activation. Surprisingly, 3D single-molecule tracking results indicate activation of the optogenetic response when illuminating with high-intensity light with wavelengths at which there is minimal photon absorbance by the LOV2 domain. The preactivation can be minimized through the use of iLID system mutants, and titration of protein expression levels.

BCM3D 2.0: accurate segmentation of single bacterial cells in dense biofilms using computationally generated intermediate image representations.

Accurate detection and segmentation of single cells in three-dimensional (3D) fluorescence time-lapse images is essential for observing individual cell behaviors in large bacterial communities called biofilms. Recent progress in machine-learning-based image analysis is providing this capability with ever-increasing accuracy. Leveraging the capabilities of deep convolutional neural networks (CNNs), we recently developed bacterial cell morphometry in 3D (BCM3D), an integrated image analysis pipeline that combines deep learning with conventional image analysis to detect and segment single biofilm-dwelling cells in 3D fluorescence images. While the first release of BCM3D (BCM3D 1.0) achieved state-of-the-art 3D bacterial cell segmentation accuracies, low signal-to-background ratios (SBRs) and images of very dense biofilms remained challenging. Here, we present BCM3D 2.0 to address this challenge. BCM3D 2.0 is entirely complementary to the approach utilized in BCM3D 1.0. Instead of training CNNs to perform voxel classification, we trained CNNs to translate 3D fluorescence images into intermediate 3D image representations that are, when combined appropriately, more amenable to conventional mathematical image processing than a single experimental image. Using this approach, improved segmentation results are obtained even for very low SBRs and/or high cell density biofilm images. The improved cell segmentation accuracies in turn enable improved accuracies of tracking individual cells through 3D space and time. This capability opens the door to investigating time-dependent phenomena in bacterial biofilms at the cellular level.

Distinct Cytosolic Complexes Containing the Type III Secretion System ATPase Resolved by Three-Dimensional Single-Molecule Tracking in Live Yersinia enterocolitica.

The membrane-embedded injectisome, the structural component of the virulence-associated type III secretion system (T3SS), is used by Gram-negative bacterial pathogens to inject species-specific effector proteins into eukaryotic host cells. The cytosolic injectisome proteins are required for export of effectors and display both stationary, injectisome-bound populations and freely diffusing cytosolic populations. How the cytosolic injectisome proteins interact with each other in the cytosol and associate with membrane-embedded injectisomes remains unclear. Here, we utilized three-dimensional (3D) single-molecule tracking to resolve distinct cytosolic complexes of injectisome proteins in living Yersinia enterocolitica cells. Tracking of the enhanced yellow fluorescent protein (eYFP)-labeled ATPase YeSctN and its regulator, YeSctL, revealed that these proteins form a cytosolic complex with each other and then further with YeSctQ. YeSctNL and YeSctNLQ complexes can be observed both in wild-type cells and in ΔsctD mutants, which cannot assemble injectisomes. In ΔsctQ mutants, the relative abundance of the YeSctNL complex is considerably increased. These data indicate that distinct cytosolic complexes of injectisome proteins can form prior to injectisome binding, which has important implications for how injectisomes are functionally regulated. IMPORTANCE Injectisomes are membrane-embedded, multiprotein assemblies used by bacterial pathogens to inject virulent effector proteins into eukaryotic host cells. Protein secretion is regulated by cytosolic proteins that dynamically bind and unbind at injectisomes. However, how these regulatory proteins interact with each other remains unknown. By measuring the diffusion rates of single molecules in living cells, we show that cytosolic injectisome proteins form distinct oligomeric complexes with each other prior to binding to injectisomes. We additionally identify the molecular compositions of these complexes and quantify their relative abundances. Quantifying to what extent cytosolic proteins exist as part of larger complexes in living cells has important implications for deciphering the complexity of biomolecular mechanisms. The results and methods reported here are thus relevant for advancing our understanding of how injectisomes and related multiprotein assemblies, such as bacterial flagellar motors, are functionally regulated.

3D GAN image synthesis and dataset quality assessment for bacterial biofilm.

MOTIVATION: Data-driven deep learning techniques usually require a large quantity of labeled training data to achieve reliable solutions in bioimage analysis. However, noisy image conditions and high cell density in bacterial biofilm images make 3D cell annotations difficult to obtain. Alternatively, data augmentation via synthetic data generation is attempted, but current methods fail to produce realistic images. RESULTS: This article presents a bioimage synthesis and assessment workflow with application to augment bacterial biofilm images. 3D cyclic generative adversarial networks (GAN) with unbalanced cycle consistency loss functions are exploited in order to synthesize 3D biofilm images from binary cell labels. Then, a stochastic synthetic dataset quality assessment (SSQA) measure that compares statistical appearance similarity between random patches from random images in two datasets is proposed. Both SSQA scores and other existing image quality measures indicate that the proposed 3D Cyclic GAN, along with the unbalanced loss function, provides a reliably realistic (as measured by mean opinion score) 3D synthetic biofilm image. In 3D cell segmentation experiments, a GAN-augmented training model also presents more realistic signal-to-background intensity ratio and improved cell counting accuracy. AVAILABILITY AND IMPLEMENTATION: https://github.com/jwang-c/DeepBiofilm. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Graph-Theoretic Post-Processing of Segmentation With Application to Dense Biofilms.

Recent deep learning methods have provided successful initial segmentation results for generalized cell segmentation in microscopy. However, for dense arrangements of small cells with limited ground truth for training, the deep learning methods produce both over-segmentation and under-segmentation errors. Post-processing attempts to balance the trade-off between the global goal of cell counting for instance segmentation, and local fidelity to the morphology of identified cells. The need for post-processing is especially evident for segmenting 3D bacterial cells in densely-packed communities called biofilms. A graph-based recursive clustering approach, m-LCuts, is proposed to automatically detect collinearly structured clusters and applied to post-process unsolved cells in 3D bacterial biofilm segmentation. Construction of outlier-removed graphs to extract the collinearity feature in the data adds additional novelty to m-LCuts. The superiority of m-LCuts is observed by the evaluation in cell counting with over 90% of cells correctly identified, while a lower bound of 0.8 in terms of average single-cell segmentation accuracy is maintained. This proposed method does not need manual specification of the number of cells to be segmented. Furthermore, the broad adaptation for working on various applications, with the presence of data collinearity, also makes m-LCuts stand out from the other approaches.

Optically Accessible Microfluidic Flow Channels for Noninvasive High-Resolution Biofilm Imaging Using Lattice Light Sheet Microscopy.

Imaging platforms that enable long-term, high-resolution imaging of biofilms are required to study cellular level dynamics within bacterial biofilms. By combining high spatial and temporal resolution and low phototoxicity, lattice light sheet microscopy (LLSM) has made critical contributions to the study of cellular dynamics. However, the power of LLSM has not yet been leveraged for biofilm research because the open-on-top imaging geometry using water-immersion objective lenses is not compatible with living bacterial specimens; bacterial growth on the microscope's objective lenses makes long-term time-lapse imaging impossible and raises considerable safety concerns for microscope users. To make LLSM compatible with pathogenic bacterial specimens, we developed hermetically sealed, but optically accessible, microfluidic flow channels that can sustain bacterial biofilm growth for multiple days under precisely controllable physical and chemical conditions. To generate a liquid- and gas-tight seal, we glued a thin polymer film across a 3D-printed channel, where the top wall had been omitted. We achieved negligible optical aberrations by using polymer films that precisely match the refractive index of water. Bacteria do not adhere to the polymer film itself, so that the polymer window provides unobstructed optical access to the channel interior. Inside the flow channels, biofilms can be grown on arbitrary, even nontransparent, surfaces. By integrating this flow channel with LLSM, we were able to record the growth of S. oneidensis MR-1 biofilms over several days at cellular resolution without any observable phototoxicity or photodamage.

Non-invasive single-cell morphometry in living bacterial biofilms.

Fluorescence microscopy enables spatial and temporal measurements of live cells and cellular communities. However, this potential has not yet been fully realized for investigations of individual cell behaviors and phenotypic changes in dense, three-dimensional (3D) bacterial biofilms. Accurate cell detection and cellular shape measurement in densely packed biofilms are challenging because of the limited resolution and low signal to background ratios (SBRs) in fluorescence microscopy images. In this work, we present Bacterial Cell Morphometry 3D (BCM3D), an image analysis workflow that combines deep learning with mathematical image analysis to accurately segment and classify single bacterial cells in 3D fluorescence images. In BCM3D, deep convolutional neural networks (CNNs) are trained using simulated biofilm images with experimentally realistic SBRs, cell densities, labeling methods, and cell shapes. We systematically evaluate the segmentation accuracy of BCM3D using both simulated and experimental images. Compared to state-of-the-art bacterial cell segmentation approaches, BCM3D consistently achieves higher segmentation accuracy and further enables automated morphometric cell classifications in multi-population biofilms.

Facile Synthesis and Metabolic Incorporation of m-DAP Bioisosteres Into Cell Walls of Live Bacteria.

Bacterial cell walls contain peptidoglycan (PG), a scaffold that provides proper rigidity to resist lysis from internal osmotic pressure and a barrier to protect cells against external stressors. It consists of repeating sugar units with a linkage to a stem peptide that becomes cross-linked by cell wall transpeptidases (TP). While synthetic PG fragments containing l-lysine in the third position on the stem peptide are easier to access, those with meso-diaminopimelic acid (m-DAP) pose a severe synthetic challenge. Herein, we describe a solid phase synthetic scheme based on widely available building blocks to assemble meso-cystine (m-CYT), which mimics key structural features of m-DAP. To demonstrate proper mimicry of m-DAP, cell wall probes were synthesized with m-CYT in place of m-DAP and evaluated for their metabolic processing in live bacterial cells. We found that m-CYT-based cell wall probes were properly processed by TPs in various bacterial species that endogenously contain m-DAP in their PG. Additionally, we have used hybrid quantum mechanical/molecular mechanical (QM/MM) and molecular dynamics (MD) simulations to explore the influence of m-DAP analogs on the PG cross-linking. The results showed that the cross-linking mechanism of transpeptidases occurred through a concerted process. We anticipate that this strategy, which is based on the use of inexpensive and commercially available building blocks, can be widely adopted to provide greater accessibility of PG mimics for m-DAP containing organisms.

Structural Similarity Image Analysis for Detection of Adenosine and Dopamine in Fast-Scan Cyclic Voltammetry Color Plots.

Fast-scan cyclic voltammetry (FSCV) is widely used for in vivo detection of neurotransmitters, but identifying analytes, particularly mixtures, is difficult. Data analysis has focused on identifying dopamine from cyclic voltammograms, but it would be better to analyze all the data in the three-dimensional FSCV color plot. Here, the goal was to use image analysis-based analysis of FSCV color plots for the first time, specifically the structural similarity index (SSIM), to identify rapid neurochemical events. Initially, we focused on identifying spontaneous adenosine events, as adenosine cyclic voltammograms have a primary oxidation at 1.3 V and a secondary oxidation peak that grows in over time. Using SSIM, sample FSCV color plots were compared with reference color plots, and the SSIM cutoff score was optimized to distinguish adenosine. High-pass digital filtering was also applied to remove the background drift and lower the noise, which produced a better LOD. The SSIM algorithm detected more adenosine events than a previous algorithm based on current versus time traces, with 99.5 ± 0.6% precision, 95 ± 3% recall, and 97 ± 2% F1 score (n = 15 experiments from three researchers). For selectivity, it successfully rejected signals from pH changes, histamine, and H2O2. To prove it is a broad strategy useful beyond adenosine, SSIM analysis was optimized for dopamine detection and is able to detect simultaneous events with dopamine and adenosine. Thus, SSIM is a general strategy for FSCV data analysis that uses three-dimensional data to detect multiple analytes in an efficient, automated analysis.

Reprogramming fatty acyl specificity of lipid kinases via C1 domain engineering.

C1 domains are lipid-binding modules that regulate membrane activation of kinases, nucleotide exchange factors and other C1-containing proteins to trigger signal transduction. Despite annotation of typical C1 domains as diacylglycerol (DAG) and phorbol ester sensors, the function of atypical counterparts remains ill-defined. Here, we assign a key role for atypical C1 domains in mediating DAG fatty acyl specificity of diacylglycerol kinases (DGKs) in live cells. Activity-based proteomics mapped C1 probe binding as a principal differentiator of type 1 DGK active sites that combined with global metabolomics revealed a role for C1s in lipid substrate recognition. Protein engineering by C1 domain swapping demonstrated that exchange of typical and atypical C1s is functionally tolerated and can directly program DAG fatty acyl specificity of type 1 DGKs. Collectively, we describe a protein engineering strategy for studying metabolic specificity of lipid kinases to assign a role for atypical C1 domains in cell metabolism.

Single-Molecule Tracking Microscopy - A Tool for Determining the Diffusive States of Cytosolic Molecules.

Single-molecule localization microscopy probes the position and motions of individual molecules in living cells with tens of nanometer spatial and millisecond temporal resolution. These capabilities make single-molecule localization microscopy ideally suited to study molecular level biological functions in physiologically relevant environments. Here, we demonstrate an integrated protocol for both acquisition and processing/analysis of single-molecule tracking data to extract the different diffusive states a protein of interest may exhibit. This information can be used to quantify molecular complex formation in living cells. We provide a detailed description of a camera-based 3D single-molecule localization experiment, as well as the subsequent data processing steps that yield the trajectories of individual molecules. These trajectories are then analyzed using a numerical analysis framework to extract the prevalent diffusive states of the fluorescently labeled molecules and the relative abundance of these states. The analysis framework is based on stochastic simulations of intracellular Brownian diffusion trajectories that are spatially confined by an arbitrary cell geometry. Based on the simulated trajectories, raw single-molecule images are generated and analyzed in the same way as experimental images. In this way, experimental precision and accuracy limitations, which are difficult to calibrate experimentally, are explicitly incorporated into the analysis workflow. The diffusion coefficient and relative population fractions of the prevalent diffusive states are determined by fitting the distributions of experimental values using linear combinations of simulated distributions. We demonstrate the utility of our protocol by resolving the diffusive states of a protein that exhibits different diffusive states upon forming homo- and hetero-oligomeric complexes in the cytosol of a bacterial pathogen.

Enabling technologies in super-resolution fluorescence microscopy: reporters, labeling, and methods of measurement.

Super-resolution fluorescence microscopy continues to experience a period of extraordinary development. New instrumentation and fluorescent labeling strategies provide access to molecular and cellular processes that occur on length scales ranging from nanometers to millimeters and on time scales ranging from milliseconds to hours. At the shortest length scales, single-molecule imaging methods now allow measurement of nanoscale localization, motion, and binding kinetics of individual biomolecules. At cellular and intercellular length scales, super-resolution microscopy allows structural and functional imaging of individual cells in tissues and even in whole animals. Here, we review recent advances that have enabled entirely new types of experiments and greatly potentiated existing technologies.

Resolving Cytosolic Diffusive States in Bacteria by Single-Molecule Tracking.

The trajectory of a single protein in the cytosol of a living cell contains information about its molecular interactions in its native environment. However, it has remained challenging to accurately resolve and characterize the diffusive states that can manifest in the cytosol using analytical approaches based on simplifying assumptions. Here, we show that multiple intracellular diffusive states can be successfully resolved if sufficient single-molecule trajectory information is available to generate well-sampled distributions of experimental measurements and if experimental biases are taken into account during data analysis. To address the inherent experimental biases in camera-based and MINFLUX-based single-molecule tracking, we use an empirical data analysis framework based on Monte Carlo simulations of confined Brownian motion. This framework is general and adaptable to arbitrary cell geometries and data acquisition parameters employed in two-dimensional or three-dimensional single-molecule tracking. We show that, in addition to determining the diffusion coefficients and populations of prevalent diffusive states, the timescales of diffusive state switching can be determined by stepwise increasing the time window of averaging over subsequent single-molecule displacements. Time-averaged diffusion analysis of single-molecule tracking data may thus provide quantitative insights into binding and unbinding reactions among rapidly diffusing molecules that are integral for cellular functions.

Single-molecule tracking in live Yersinia enterocolitica reveals distinct cytosolic complexes of injectisome subunits.

In bacterial type 3 secretion, substrate proteins are actively transported from the bacterial cytoplasm into the host cell cytoplasm by a large membrane-embedded machinery called the injectisome. Injectisomes transport secretion substrates in response to specific environmental signals, but the molecular details by which the cytosolic secretion substrates are selected and transported through the type 3 secretion pathway remain unclear. Secretion activity and substrate selectivity are thought to be controlled by a sorting platform consisting of the proteins SctK, SctQ, SctL, and SctN, which together localize to the cytoplasmic side of membrane-embedded injectisomes. However, recent work revealed that sorting platform proteins additionally exhibit substantial cytosolic populations and that SctQ reversibly binds to and dissociates from the cytoplasmic side of membrane-embedded injectisomes. Based on these observations, we hypothesized that dynamic molecular turnover at the injectisome and cytosolic assembly among sorting platform proteins is a critical regulatory component of type 3 secretion. To determine whether sorting platform complexes exist in the cytosol, we measured the diffusive properties of the two central sorting platform proteins, SctQ and SctL, using live cell high-throughput 3D single-molecule tracking microscopy. Single-molecule trajectories, measured in wild-type and mutant Yersinia enterocolitica cells, reveal that both SctQ and SctL exist in several distinct diffusive states in the cytosol, indicating that these proteins form stable homo- and hetero-oligomeric complexes in their native environment. Our findings provide the first diffusive state-resolved insights into the dynamic regulatory network that interfaces stationary membrane-embedded injectisomes with the soluble cytosolic components of the type 3 secretion system.

Bacterial scaffold directs pole-specific centromere segregation.

Bacteria use partitioning systems based on the ParA ATPase to actively mobilize and spatially organize molecular cargoes throughout the cytoplasm. The bacterium Caulobacter crescentus uses a ParA-based partitioning system to segregate newly replicated chromosomal centromeres to opposite cell poles. Here we demonstrate that the Caulobacter PopZ scaffold creates an organizing center at the cell pole that actively regulates polar centromere transport by the ParA partition system. As segregation proceeds, the ParB-bound centromere complex is moved by progressively disassembling ParA from a nucleoid-bound structure. Using superresolution microscopy, we show that released ParA is recruited directly to binding sites within a 3D ultrastructure composed of PopZ at the cell pole, whereas the ParB-centromere complex remains at the periphery of the PopZ structure. PopZ recruitment of ParA stimulates ParA to assemble on the nucleoid near the PopZ-proximal cell pole. We identify mutations in PopZ that allow scaffold assembly but specifically abrogate interactions with ParA and demonstrate that PopZ/ParA interactions are required for proper chromosome segregation in vivo. We propose that during segregation PopZ sequesters free ParA and induces target-proximal regeneration of ParA DNA binding activity to enforce processive and pole-directed centromere segregation, preventing segregation reversals. PopZ therefore functions as a polar hub complex at the cell pole to directly regulate the directionality and destination of transfer of the mitotic segregation machine.

Exploring bacterial cell biology with single-molecule tracking and super-resolution imaging.

The ability to detect single molecules in live bacterial cells enables us to probe biological events one molecule at a time and thereby gain knowledge of the activities of intracellular molecules that remain obscure in conventional ensemble-averaged measurements. Single-molecule fluorescence tracking and super-resolution imaging are thus providing a new window into bacterial cells and facilitating the elucidation of cellular processes at an unprecedented level of sensitivity, specificity and spatial resolution. In this Review, we consider what these technologies have taught us about the bacterial cytoskeleton, nucleoid organization and the dynamic processes of transcription and translation, and we also highlight the methodological improvements that are needed to address a number of experimental challenges in the field.

Quantitative multicolor subdiffraction imaging of bacterial protein ultrastructures in three dimensions.

We demonstrate quantitative multicolor three-dimensional (3D) subdiffraction imaging of the structural arrangement of fluorescent protein fusions in living Caulobacter crescentus bacteria. Given single-molecule localization precisions of 20-40 nm, a flexible locally weighted image registration algorithm is critical to accurately combine the super-resolution data with <10 nm error. Surface-relief dielectric phase masks implement a double-helix response at two wavelengths to distinguish two different fluorescent labels and to quantitatively and precisely localize them relative to each other in 3D.