Imaging Flow Cytometry and Convolutional Neural Network-Based Classification Enable Discrimination of Hematopoietic and Leukemic Stem Cells in Acute Myeloid Leukemia.

Acute myeloid leukemia (AML) is a heterogenous blood cancer with a dismal prognosis. It emanates from leukemic stem cells (LSCs) arising from the genetic transformation of hematopoietic stem cells (HSCs). LSCs hold prognostic value, but their molecular and immunophenotypic heterogeneity poses challenges: there is no single marker for identifying all LSCs across AML samples. We hypothesized that imaging flow cytometry (IFC) paired with artificial intelligence-driven image analysis could visually distinguish LSCs from HSCs based solely on morphology. Initially, a seven-color IFC panel was employed to immunophenotypically identify LSCs and HSCs in bone marrow samples from five AML patients and ten healthy donors, respectively. Next, we developed convolutional neural network (CNN) models for HSC-LSC discrimination using brightfield (BF), side scatter (SSC), and DNA images. Classification using only BF images achieved 86.96% accuracy, indicating significant morphological differences. Accuracy increased to 93.42% when combining BF with DNA images, highlighting differences in nuclear morphology, although DNA images alone were inadequate for accurate HSC-LSC discrimination. Model development using SSC images revealed minor granularity differences. Performance metrics varied substantially between AML patients, indicating considerable morphologic variations among LSCs. Overall, we demonstrate proof-of-concept results for accurate CNN-based HSC-LSC differentiation, instigating the development of a novel technique within AML monitoring.

Comparative analysis of feature-based ML and CNN for binucleated erythroblast quantification in myelodysplastic syndrome patients using imaging flow cytometry data.

Myelodysplastic syndrome is primarily characterized by dysplasia in the bone marrow (BM), presenting a challenge in consistent morphology interpretation. Accurate diagnosis through traditional slide-based analysis is difficult, necessitating a standardized objective technique. Over the past two decades, imaging flow cytometry (IFC) has proven effective in combining image-based morphometric analyses with high-parameter phenotyping. We have previously demonstrated the effectiveness of combining IFC with a feature-based machine learning algorithm to accurately identify and quantify rare binucleated erythroblasts (BNEs) in dyserythropoietic BM cells. However, a feature-based workflow poses challenges requiring software-specific expertise. Here we employ a Convolutional Neural Network (CNN) algorithm for BNE identification and differentiation from doublets and cells with irregular nuclear morphology in IFC data. We demonstrate that this simplified AI workflow, coupled with a powerful CNN algorithm, achieves comparable BNE quantification accuracy to manual and feature-based analysis with substantial time savings, eliminating workflow complexity. This streamlined approach holds significant clinical value, enhancing IFC accessibility for routine diagnostic purposes.

Graph-based algorithms to dissect long-distance water-mediated H-bond networks for conformational couplings in GPCRs.

Changes in structure and dynamics elicited by agonist ligand binding at the extracellular side of G protein coupled receptors (GPCRs) must be relayed to the cytoplasmic G protein binding side of the receptors. To decipher the role of water-mediated hydrogen-bond networks in this relay mechanism, we have developed graph-based algorithms and analysis methodologies applicable to datasets of static structures of distinct GPCRs. For a reference dataset of static structures of bovine rhodopsin solved at the same resolution, we show that graph analyses capture the internal protein-water hydrogen-bond network. The extended analyses of static structures of rhodopsins and opioid receptors suggest a relay mechanism whereby inactive receptors have in place much of the internal core hydrogen-bond network required for long-distance relay of structural change, with extensive local H-bond clusters observed in structures solved at high resolution and with internal water molecules.

A redox switch allows binding of Fe(II) and Fe(III) ions in the cyanobacterial iron-binding protein FutA from Prochlorococcus.

The marine cyanobacterium Prochlorococcus is a main contributor to global photosynthesis, whilst being limited by iron availability. Cyanobacterial genomes generally encode two different types of FutA iron-binding proteins: periplasmic FutA2 ABC transporter subunits bind Fe(III), while cytosolic FutA1 binds Fe(II). Owing to their small size and their economized genome Prochlorococcus ecotypes typically possess a single futA gene. How the encoded FutA protein might bind different Fe oxidation states was previously unknown. Here, we use structural biology techniques at room temperature to probe the dynamic behavior of FutA. Neutron diffraction confirmed four negatively charged tyrosinates, that together with a neutral water molecule coordinate iron in trigonal bipyramidal geometry. Positioning of the positively charged Arg103 side chain in the second coordination shell yields an overall charge-neutral Fe(III) binding state in structures determined by neutron diffraction and serial femtosecond crystallography. Conventional rotation X-ray crystallography using a home source revealed X-ray-induced photoreduction of the iron center with observation of the Fe(II) binding state; here, an additional positioning of the Arg203 side chain in the second coordination shell maintained an overall charge neutral Fe(II) binding site. Dose series using serial synchrotron crystallography and an XFEL X-ray pump-probe approach capture the transition between Fe(III) and Fe(II) states, revealing how Arg203 operates as a switch to accommodate the different iron oxidation states. This switching ability of the Prochlorococcus FutA protein may reflect ecological adaptation by genome streamlining and loss of specialized FutA proteins.

Application of the Cytokinesis-Block Micronucleus Assay for High-Dose Exposures Using Imaging Flow Cytometry.

The cytokinesis-block micronucleus assay is a well-established method to assess radiation-induced genetic damage in human cells. This assay has been adapted to imaging flow cytometry (IFC), allowing automated analysis of many cells, and eliminating the need to create microscope slides. Furthermore, to improve the efficiency of assay performance, a small-volume method previously developed was employed. Irradiated human blood samples were cultured, stained, and analyzed by IFC to produce images of the cells. Samples were run using both manual and 96-well plate automated acquisition. Multiple parameter-based image features were collected for each sample, and the results were compared to confirm that these acquisition methods are functionally identical. This paper details the multi-parametric analysis developed and the resulting calibration curves up to 10 Gy. The calibration curves were created using a quadratic random coefficient model with Poisson errors, as well as a logistic discriminant function. The curves were then validated with blinded, irradiated samples, using relative bias and relative mean square error. Overall, the accuracy of the dose estimates was adequate for triage dosimetry (within 1 Gy of the true dose) over 90% of the time for lower doses and about half the time for higher doses, with the lowest success rate between 5 and 6 Gy where the calibration curve reached its peak and there was the smallest change in MN/BNC with dose. This work describes the application of a novel multi-parametric analysis that fits the calibration curves and allows dose estimates up to 10 Gy, which were previously limited to 4 Gy. Furthermore, it demonstrates that the results from samples acquired manually and with the autosampler are functionally similar.

Pyroptosis Markers in Human Primary Specimens: Quantification of Intracellular ASC Specks by Imaging Flow Cytometry and Extracellular Oxidized Mitochondrial by ELISA.

Pyroptosis is an immunological response to infection and cellular stresses initiated by inflammasome oligomerization resulting in the release of pro-inflammatory factors including cytokines and other immune stimuli into the extracellular matrix. In order to understand the role of inflammasome activation and subsequent pyroptosis in human infection and disease pathogenesis and to explore markers of these signaling events as potential disease or response biomarkers, we must utilize quantitative, reliable, and reproducible assays to readily investigate these pathways in primary specimens. Here, we describe two methods using imaging flow cytometry for evaluation of inflammasome ASC specks in homogeneous peripheral blood monocytes and in bulk, heterogeneous peripheral blood mononuclear cells. Both methods can be applied to assess speck formation as a biomarker for inflammasome activation in primary specimens. Additionally, we describe the methods for quantification of extracellular oxidized mitochondrial DNA from primary plasma samples, serving as a proxy for pyroptosis. Collectively, these assays may be utilized to determine pyroptotic influences on viral infection and disease development or as diagnostic aids and response biomarkers.

The Imaging Flow Cytometry-Based Cytokinesis-Block MicroNucleus (CBMN) Assay.

The dose of ionizing radiation received by an individual can be determined using biodosimetry methods which measure biomarkers of exposure in tissue samples from that individual. These markers can be expressed in many ways, including DNA damage and repair processes. Following a mass casualty event involving radiological or nuclear material, it is important to rapidly provide this information to medical responders to assist in the medical management of potentially exposed casualties. Traditional methods of biodosimetry rely on microscope analysis, making them time-consuming and labor-intensive. To increase sample throughput following a large-scale radiological mass casualty event, several biodosimetry assays have been adapted for analysis by imaging flow cytometry. This chapter briefly reviews these methods with a focus on the most current methodology to identify and quantify micronuclei in binucleated cells within the cytokinesis-block micronucleus assay using an imaging flow cytometer.

Structural basis of calmodulin modulation of the rod cyclic nucleotide-gated channel.

Calmodulin (CaM) regulates many ion channels to control calcium entry into cells, and mutations that alter this interaction are linked to fatal diseases. The structural basis of CaM regulation remains largely unexplored. In retinal photoreceptors, CaM binds to the CNGB subunit of cyclic nucleotide-gated (CNG) channels and, thereby, adjusts the channel's Cyclic guanosine monophosphate (cGMP) sensitivity in response to changes in ambient light conditions. Here, we provide the structural characterization for CaM regulation of a CNG channel by using a combination of single-particle cryo-electron microscopy and structural proteomics. CaM connects the CNGA and CNGB subunits, resulting in structural changes both in the cytosolic and transmembrane regions of the channel. Cross-linking and limited proteolysis-coupled mass spectrometry mapped the conformational changes induced by CaM in vitro and in the native membrane. We propose that CaM is a constitutive subunit of the rod channel to ensure high sensitivity in dim light. Our mass spectrometry-based approach is generally relevant for studying the effect of CaM on ion channels in tissues of medical interest, where only minute quantities are available.

Correction of rhodopsin serial crystallography diffraction intensities for a lattice-translocation defect.

Rhodopsin is a G-protein-coupled receptor that detects light and initiates the intracellular signalling cascades that underpin vertebrate vision. Light sensitivity is achieved by covalent linkage to 11-cis retinal, which isomerizes upon photo-absorption. Serial femtosecond crystallography data collected from rhodopsin microcrystals grown in the lipidic cubic phase were used to solve the room-temperature structure of the receptor. Although the diffraction data showed high completeness and good consistency to 1.8 Šresolution, prominent electron-density features remained unaccounted for throughout the unit cell after model building and refinement. A deeper analysis of the diffraction intensities uncovered the presence of a lattice-translocation defect (LTD) within the crystals. The procedure followed to correct the diffraction intensities for this pathology enabled the building of an improved resting-state model. The correction was essential to both confidently model the structure of the unilluminated state and interpret the light-activated data collected after photo-excitation of the crystals. It is expected that similar cases of LTD will be observed in other serial crystallography experiments and that correction will be required in a variety of systems.

Ultrafast structural changes direct the first molecular events of vision.

Vision is initiated by the rhodopsin family of light-sensitive G protein-coupled receptors (GPCRs)1. A photon is absorbed by the 11-cis retinal chromophore of rhodopsin, which isomerizes within 200 femtoseconds to the all-trans conformation2, thereby initiating the cellular signal transduction processes that ultimately lead to vision. However, the intramolecular mechanism by which the photoactivated retinal induces the activation events inside rhodopsin remains experimentally unclear. Here we use ultrafast time-resolved crystallography at room temperature3 to determine how an isomerized twisted all-trans retinal stores the photon energy that is required to initiate the protein conformational changes associated with the formation of the G protein-binding signalling state. The distorted retinal at a 1-ps time delay after photoactivation has pulled away from half of its numerous interactions with its binding pocket, and the excess of the photon energy is released through an anisotropic protein breathing motion in the direction of the extracellular space. Notably, the very early structural motions in the protein side chains of rhodopsin appear in regions that are involved in later stages of the conserved class A GPCR activation mechanism. Our study sheds light on the earliest stages of vision in vertebrates and points to fundamental aspects of the molecular mechanisms of agonist-mediated GPCR activation.

Automation of the Micronucleus Assay Using Imaging Flow Cytometry and Artificial Intelligence.

The micronucleus (MN) assay is used worldwide by regulatory bodies to evaluate chemicals for genetic toxicity. The assay can be performed in two ways: by scoring MN in once-divided, cytokinesis-blocked binucleated cells or fully divided mononucleated cells. Historically, light microscopy has been the gold standard method to score the assay, but it is laborious and subjective. Flow cytometry has been used in recent years to score the assay, but is limited by the inability to visually confirm key aspects of cellular imagery. Imaging flow cytometry (IFC) combines high-throughput image capture and automated image analysis, and has been successfully applied to rapidly acquire imagery of and score all key events in the MN assay. Recently, it has been demonstrated that artificial intelligence (AI) methods based on convolutional neural networks can be used to score MN assay data acquired by IFC. This paper describes all steps to use AI software to create a deep learning model to score all key events and to apply this model to automatically score additional data. Results from the AI deep learning model compare well to manual microscopy, therefore enabling fully automated scoring of the MN assay by combining IFC and AI.

Discovering cell-active BCL6 inhibitors: effectively combining biochemical HTS with multiple biophysical techniques, X-ray crystallography and cell-based assays.

By suppressing gene transcription through the recruitment of corepressor proteins, B-cell lymphoma 6 (BCL6) protein controls a transcriptional network required for the formation and maintenance of B-cell germinal centres. As BCL6 deregulation is implicated in the development of Diffuse Large B-Cell Lymphoma, we sought to discover novel small molecule inhibitors that disrupt the BCL6-corepressor protein-protein interaction (PPI). Here we report our hit finding and compound optimisation strategies, which provide insight into the multi-faceted orthogonal approaches that are needed to tackle this challenging PPI with small molecule inhibitors. Using a 1536-well plate fluorescence polarisation high throughput screen we identified multiple hit series, which were followed up by hit confirmation using a thermal shift assay, surface plasmon resonance and ligand-observed NMR. We determined X-ray structures of BCL6 bound to compounds from nine different series, enabling a structure-based drug design approach to improve their weak biochemical potency. We developed a time-resolved fluorescence energy transfer biochemical assay and a nano bioluminescence resonance energy transfer cellular assay to monitor cellular activity during compound optimisation. This workflow led to the discovery of novel inhibitors with respective biochemical and cellular potencies (IC50s) in the sub-micromolar and low micromolar range.

Somatic gene mutations expose cytoplasmic DNA to co-opt the cGAS/STING/NLRP3 axis in myelodysplastic syndromes.

NLRP3 inflammasome and IFN-stimulated gene (ISG) induction are key biological drivers of ineffective hematopoiesis and inflammation in myelodysplastic syndromes (MDSs). Gene mutations involving mRNA splicing and epigenetic regulatory pathways induce inflammasome activation and myeloid lineage skewing in MDSs through undefined mechanisms. Using immortalized murine hematopoietic stem and progenitor cells harboring these somatic gene mutations and primary MDS BM specimens, we showed accumulation of unresolved R-loops and micronuclei with concurrent activation of the cytosolic sensor cyclic GMP-AMP synthase. Cyclic GMP-AMP synthase/stimulator of IFN genes (cGAS/STING) signaling caused ISG induction, NLRP3 inflammasome activation, and maturation of the effector protease caspase-1. Deregulation of RNA polymerase III drove cytosolic R-loop generation, which upon inhibition, extinguished ISG and inflammasome response. Mechanistically, caspase-1 degraded the master erythroid transcription factor, GATA binding protein 1, provoking anemia and myeloid lineage bias that was reversed by cGAS inhibition in vitro and in Tet2-/- hematopoietic stem and progenitor cell-transplanted mice. Together, these data identified a mechanism by which functionally distinct mutations converged upon the cGAS/STING/NLRP3 axis in MDS, directing ISG induction, pyroptosis, and myeloid lineage skewing.

Optimizing Shape Complementarity Enables the Discovery of Potent Tricyclic BCL6 Inhibitors.

To identify new chemical series with enhanced binding affinity to the BTB domain of B-cell lymphoma 6 protein, we targeted a subpocket adjacent to Val18. With no opportunities for strong polar interactions, we focused on attaining close shape complementarity by ring fusion onto our quinolinone lead series. Following exploration of different sized rings, we identified a conformationally restricted core which optimally filled the available space, leading to potent BCL6 inhibitors. Through X-ray structure-guided design, combined with efficient synthetic chemistry to make the resulting novel core structures, a >300-fold improvement in activity was obtained by the addition of seven heavy atoms.

Trapping and structural characterisation of a covalent intermediate in vitamin B6 biosynthesis catalysed by the Pdx1 PLP synthase.

The Pdx1 enzyme catalyses condensation of two carbohydrates and ammonia to form pyridoxal 5-phosphate (PLP) via an imine relay mechanism of carbonyl intermediates. The I333 intermediate characterised here using structural, UV-vis absorption spectroscopy and mass spectrometry analyses rationalises stereoselective deprotonation and subsequent substrate assisted phosphate elimination, central to PLP biosynthesis.

Into Deep Water: Optimizing BCL6 Inhibitors by Growing into a Solvated Pocket.

We describe the optimization of modestly active starting points to potent inhibitors of BCL6 by growing into a subpocket, which was occupied by a network of five stably bound water molecules. Identifying potent inhibitors required not only forming new interactions in the subpocket but also perturbing the water network in a productive, potency-increasing fashion while controlling the physicochemical properties. We achieved this goal in a sequential manner by systematically probing the pocket and the water network, ultimately achieving a 100-fold improvement of activity. The most potent compounds displaced three of the five initial water molecules and formed hydrogen bonds with the remaining two. Compound 25 showed a promising profile for a lead compound with submicromolar inhibition of BCL6 in cells and satisfactory pharmacokinetic (PK) properties. Our work highlights the importance of finding productive ways to perturb existing water networks when growing into solvent-filled protein pockets.

Inter-laboratory automation of the in vitro micronucleus assay using imaging flow cytometry and deep learning.

The in vitro micronucleus assay is a globally significant method for DNA damage quantification used for regulatory compound safety testing in addition to inter-individual monitoring of environmental, lifestyle and occupational factors. However, it relies on time-consuming and user-subjective manual scoring. Here we show that imaging flow cytometry and deep learning image classification represents a capable platform for automated, inter-laboratory operation. Images were captured for the cytokinesis-block micronucleus (CBMN) assay across three laboratories using methyl methanesulphonate (1.25-5.0 µg/mL) and/or carbendazim (0.8-1.6 µg/mL) exposures to TK6 cells. Human-scored image sets were assembled and used to train and test the classification abilities of the "DeepFlow" neural network in both intra- and inter-laboratory contexts. Harnessing image diversity across laboratories yielded a network able to score unseen data from an entirely new laboratory without any user configuration. Image classification accuracies of 98%, 95%, 82% and 85% were achieved for 'mononucleates', 'binucleates', 'mononucleates with MN' and 'binucleates with MN', respectively. Successful classifications of 'trinucleates' (90%) and 'tetranucleates' (88%) in addition to 'other or unscorable' phenotypes (96%) were also achieved. Attempts to classify extremely rare, tri- and tetranucleated cells with micronuclei into their own categories were less successful (≤ 57%). Benchmark dose analyses of human or automatically scored micronucleus frequency data yielded quantitation of the same equipotent concentration regardless of scoring method. We conclude that this automated approach offers significant potential to broaden the practical utility of the CBMN method across industry, research and clinical domains. We share our strategy using openly-accessible frameworks.

The in vitro micronucleus assay using imaging flow cytometry and deep learning.

The in vitro micronucleus (MN) assay is a well-established assay for quantification of DNA damage, and is required by regulatory bodies worldwide to screen chemicals for genetic toxicity. The MN assay is performed in two variations: scoring MN in cytokinesis-blocked binucleated cells or directly in unblocked mononucleated cells. Several methods have been developed to score the MN assay, including manual and automated microscopy, and conventional flow cytometry, each with advantages and limitations. Previously, we applied imaging flow cytometry (IFC) using the ImageStream® to develop a rapid and automated MN assay based on high throughput image capture and feature-based image analysis in the IDEAS® software. However, the analysis strategy required rigorous optimization across chemicals and cell lines. To overcome the complexity and rigidity of feature-based image analysis, in this study we used the Amnis® AI software to develop a deep-learning method based on convolutional neural networks to score IFC data in both the cytokinesis-blocked and unblocked versions of the MN assay. We show that the use of the Amnis AI software to score imagery acquired using the ImageStream® compares well to manual microscopy and outperforms IDEAS® feature-based analysis, facilitating full automation of the MN assay.

The 3D reconstructed skin micronucleus assay using imaging flow cytometry and deep learning: A proof-of-principle investigation.

The reconstructed skin micronucleus (RSMN) assay was developed in 2006, as an in vitro alternative for genotoxicity evaluation of dermally applied chemicals or products. In the years since, significant progress has been made in the optimization of the assay, including publication of a standard protocol and extensive validation. However, the diverse morphology of skin cells makes cell preparation and scoring of micronuclei (MN) tedious and subjective, thus requiring a high level of technical expertise for evaluation. This ultimately has a negative impact on throughput and the assay would benefit by the development of an automated method which could reduce scoring subjectivity while also improving the robustness of the assay by increasing the number of cells that can be scored. Imaging flow cytometry (IFC) with the ImageStream®X Mk II can capture high-resolution transmission and fluorescent imagery of cells in suspension. This proof-of-principle study describes protocol modifications that enable such automated measurement in 3D skin cells following exposure to mitomycin C and colchicine. IFC was then used for automated image capture and the Amnis® Artificial Intelligence (AAI) software permitted identification of binucleated (BN) cells with 91% precision. On average, three times as many BN cells from control samples were evaluated using IFC compared to the standard manual analysis. When IFC MNBN cells were visually scored from within the BN cell images, their frequency compared well with manual slide scoring, showing that IFC technology can be applied to the RSMN assay. This method enables faster time to result than microscope-based scoring and the initial studies presented here demonstrate its capability for the detection of statistically significant increases in MNBN frequencies. This work therefore demonstrates the feasibility of combining IFC and AAI to automate scoring for the RSMN assay and to improve its throughput and statistical robustness.