Is AI essential? Examining the need for deep learning in image-activated sorting of Saccharomyces cerevisiae.

Artificial intelligence (AI) has become a focal point across a multitude of societal sectors, with science not being an exception. Particularly in the life sciences, imaging flow cytometry has increasingly integrated AI for automated management and categorization of extensive cell image data. However, the necessity of AI over traditional classification methods when extending imaging flow cytometry to include cell sorting remains uncertain, primarily due to the time constraints between image acquisition and sorting actuation. AI-enabled image-activated cell sorting (IACS) methods remain substantially limited, even as recent advancements in IACS have found success while largely relying on traditional feature gating strategies. Here we assess the necessity of AI for image classification in IACS by contrasting the performance of feature gating, classical machine learning (ML), and deep learning (DL) with convolutional neural networks (CNNs) in the differentiation of Saccharomyces cerevisiae mutant images. We show that classical ML could only yield a 2.8-fold enhancement in target enrichment capability, albeit at the cost of a 13.7-fold increase in processing time. Conversely, a CNN could offer an 11.0-fold improvement in enrichment capability at an 11.5-fold increase in processing time. We further executed IACS on mixed mutant populations and quantified target strain enrichment via downstream DNA sequencing to substantiate the applicability of DL for the proposed study. Our findings validate the feasibility and value of employing DL in IACS for morphology-based genetic screening of S. cerevisiae, encouraging its incorporation in future advancements of similar technologies.

Real-time intelligent classification of COVID-19 and thrombosis via massive image-based analysis of platelet aggregates.

Microvascular thrombosis is a typical symptom of COVID-19 and shows similarities to thrombosis. Using a microfluidic imaging flow cytometer, we measured the blood of 181 COVID-19 samples and 101 non-COVID-19 thrombosis samples, resulting in a total of 6.3 million bright-field images. We trained a convolutional neural network to distinguish single platelets, platelet aggregates, and white blood cells and performed classical image analysis for each subpopulation individually. Based on derived single-cell features for each population, we trained machine learning models for classification between COVID-19 and non-COVID-19 thrombosis, resulting in a patient testing accuracy of 75%. This result indicates that platelet formation differs between COVID-19 and non-COVID-19 thrombosis. All analysis steps were optimized for efficiency and implemented in an easy-to-use plugin for the image viewer napari, allowing the entire analysis to be performed within seconds on mid-range computers, which could be used for real-time diagnosis.

Image-based cell sorting using focused travelling surface acoustic waves.

Sorting cells is an essential primary step in many biological and clinical applications such as high-throughput drug screening, cancer research and cell transplantation. Cell sorting based on their mechanical properties has long been considered as a promising label-free biomarker that could revolutionize the isolation of cells from heterogeneous populations. Recent advances in microfluidic image-based cell analysis combined with subsequent label-free sorting by on-chip actuators demonstrated the possibility of sorting cells based on their physical properties. However, the high purity of sorting is achieved at the expense of a sorting rate that lags behind the analysis throughput. Furthermore, stable and reliable system operation is an important feature in enabling the sorting of small cell fractions from a concentrated heterogeneous population. Here, we present a label-free cell sorting method, based on the use of focused travelling surface acoustic wave (FTSAW) in combination with real-time deformability cytometry (RT-DC). We demonstrate the flexibility and applicability of the method by sorting distinct blood cell types, cell lines and particles based on different physical parameters. Finally, we present a new strategy to sort cells based on their mechanical properties. Our system enables the sorting of up to 400 particles per s. Sorting is therefore possible at high cell concentrations (up to 36 million per ml) while retaining high purity (>92%) for cells with diverse sizes and mechanical properties moving in a highly viscous buffer. Sorting of small cell fraction from a heterogeneous population prepared by processing of small sample volume (10 µl) is also possible and here demonstrated by the 667-fold enrichment of white blood cells (WBCs) from raw diluted whole blood in a continuous 10-hour sorting experiment. The real-time analysis of multiple parameters together with the high sensitivity and high-throughput of our method thus enables new biological and therapeutic applications in the future.

Intelligent sort-timing prediction for image-activated cell sorting.

Intelligent image-activated cell sorting (iIACS) has enabled high-throughput image-based sorting of single cells with artificial intelligence (AI) algorithms. This AI-on-a-chip technology combines fluorescence microscopy, AI-based image processing, sort-timing prediction, and cell sorting. Sort-timing prediction is particularly essential due to the latency on the order of milliseconds between image acquisition and sort actuation, during which image processing is performed. The long latency amplifies the effects of the fluctuations in the flow speed of cells, leading to fluctuation and uncertainty in the arrival time of cells at the sort point on the microfluidic chip. To compensate for this fluctuation, iIACS measures the flow speed of each cell upstream, predicts the arrival timing of the cell at the sort point, and activates the actuation of the cell sorter appropriately. Here, we propose and demonstrate a machine learning technique to increase the accuracy of the sort-timing prediction that would allow for the improvement of sort event rate, yield, and purity. Specifically, we trained an algorithm to predict the sort timing for morphologically heterogeneous budding yeast cells. The algorithm we developed used cell morphology, position, and flow speed as inputs for prediction and achieved 41.5% lower prediction error compared to the previously employed method based solely on flow speed. As a result, our technique would allow for an increase in the sort event rate of iIACS by a factor of ~2.

Long-term effects of Pfizer-BioNTech COVID-19 vaccinations on platelets.

There is a global concern about the safety of COVID-19 vaccines associated with platelet function. However, their long-term effects on overall platelet activity remain poorly understood. Here we address this problem by image-based single-cell profiling and temporal monitoring of circulating platelet aggregates in the blood of healthy human subjects, before and after they received multiple Pfizer-BioNTech (BNT162b2) vaccine doses over a time span of nearly 1 year. Results show no significant or persisting platelet aggregation trends following the vaccine doses, indicating that any effects of vaccinations on platelet turnover, platelet activation, platelet aggregation, and platelet-leukocyte interaction was insignificant.

Perturbations of mesenchymal stromal cells after allogeneic hematopoietic cell transplantation predispose for bone marrow graft-versus-host-disease.

Functional impairment of the bone marrow (BM) niche has been suggested as a major reason for prolonged cytopenia and secondary graft failure after allogeneic hematopoietic cell transplantation (alloHCT). Because mesenchymal stromal cells (MSCs) serve as multipotent progenitors for several niche components in the BM, they might play a key role in this process. We used collagenase digested trephine biopsies to directly quantify MSCs in 73 patients before (n = 18) and/or after alloHCT (n = 65). For the first time, we demonstrate that acute graft-versus-host disease (aGvHD, n = 39) is associated with a significant decrease in MSC numbers. MSC reduction can be observed even before the clinical onset of aGvHD (n = 10). Assessing MSCs instantly after biopsy collection revealed phenotypic and functional differences depending on the occurrence of aGvHD. These differences vanished during ex vivo expansion. The MSC endotypes observed revealed an enhanced population of donor-derived classical dendritic cells type 1 and alloreactive T cells as the causing agent for compartmental inflammation and MSC damage before clinical onset of aGvHD was ascertained. In conclusion, MSCs endotypes may constitute a predisposing conductor of alloreactivity after alloHCT preceding the clinical diagnosis of aGvHD.

Intelligent image-activated sorting of Chlamydomonas reinhardtii by mitochondrial localization.

Organelle positioning in cells is associated with various metabolic functions and signaling in unicellular organisms. Specifically, the microalga Chlamydomonas reinhardtii repositions its mitochondria, depending on the levels of inorganic carbon. Mitochondria are typically randomly distributed in the Chlamydomonas cytoplasm, but relocate toward the cell periphery at low inorganic carbon levels. This mitochondrial relocation is linked with the carbon-concentrating mechanism, but its significance is not yet thoroughly understood. A genotypic understanding of this relocation would require a high-throughput method to isolate rare mutant cells not exhibiting this relocation. However, this task is technically challenging due to the complex intracellular morphological difference between mutant and wild-type cells, rendering conventional non-image-based high-event-rate methods unsuitable. Here, we report our demonstration of intelligent image-activated cell sorting by mitochondrial localization. Specifically, we applied an intelligent image-activated cell sorting system to sort for C. reinhardtii cells displaying no mitochondrial relocation. We trained a convolutional neural network (CNN) to distinguish the cell types based on the complex morphology of their mitochondria. The CNN was employed to perform image-activated sorting for the mutant cell type at 180 events per second, which is 1-2 orders of magnitude faster than automated microscopy with robotic pipetting, resulting in an enhancement of the concentration from 5% to 56.5% corresponding to an enrichment factor of 11.3. These results show the potential of image-activated cell sorting for connecting genotype-phenotype relations for rare-cell populations, which require a high throughput and could lead to a better understanding of metabolic functions in cells.

Changes in Blood Cell Deformability in Chorea-Acanthocytosis and Effects of Treatment With Dasatinib or Lithium.

Misshaped red blood cells (RBCs), characterized by thorn-like protrusions known as acanthocytes, are a key diagnostic feature in Chorea-Acanthocytosis (ChAc), a rare neurodegenerative disorder. The altered RBC morphology likely influences their biomechanical properties which are crucial for the cells to pass the microvasculature. Here, we investigated blood cell deformability of five ChAc patients compared to healthy controls during up to 1-year individual off-label treatment with the tyrosine kinase inhibitor dasatinib or several weeks with lithium. Measurements with two microfluidic techniques allowed us to assess RBC deformability under different shear stresses. Furthermore, we characterized leukocyte stiffness at high shear stresses. The results showed that blood cell deformability-including both RBCs and leukocytes - in general was altered in ChAc patients compared to healthy donors. Therefore, this study shows for the first time an impairment of leukocyte properties in ChAc. During treatment with dasatinib or lithium, we observed alterations in RBC deformability and a stiffness increase for leukocytes. The hematological phenotype of ChAc patients hinted at a reorganization of the cytoskeleton in blood cells which partly explains the altered mechanical properties observed here. These findings highlight the need for a systematic assessment of the contribution of impaired blood cell mechanics to the clinical manifestation of ChAc.

Depressive disorders are associated with increased peripheral blood cell deformability: a cross-sectional case-control study (Mood-Morph).

Pathophysiological landmarks of depressive disorders are chronic low-grade inflammation and elevated glucocorticoid output. Both can potentially interfere with cytoskeleton organization, cell membrane bending and cell function, suggesting altered cell morpho-rheological properties like cell deformability and other cell mechanical features in depressive disorders. We performed a cross-sectional case-control study using the image-based morpho-rheological characterization of unmanipulated blood samples facilitating real-time deformability cytometry (RT-DC). Sixty-nine pre-screened individuals at high risk for depressive disorders and 70 matched healthy controls were included and clinically evaluated by Composite International Diagnostic Interview leading to lifetime and 12-month diagnoses. Facilitating deep learning on blood cell images, major blood cell types were classified and morpho-rheological parameters such as cell size and cell deformability of every individual cell was quantified. We found peripheral blood cells to be more deformable in patients with depressive disorders compared to controls, while cell size was not affected. Lifetime persistent depressive disorder was associated with increased cell deformability in monocytes and neutrophils, while in 12-month persistent depressive disorder erythrocytes deformed more. Lymphocytes were more deformable in 12-month major depressive disorder, while for lifetime major depressive disorder no differences could be identified. After correction for multiple testing, only associations for lifetime persistent depressive disorder remained significant. This is the first study analyzing morpho-rheological properties of entire blood cells and highlighting depressive disorders and in particular persistent depressive disorders to be associated with increased blood cell deformability. While all major blood cells tend to be more deformable, lymphocytes, monocytes, and neutrophils are mostly affected. This indicates that immune cell mechanical changes occur in depressive disorders, which might be predictive of persistent immune response.

Deep imaging flow cytometry.

Imaging flow cytometry (IFC) has become a powerful tool for diverse biomedical applications by virtue of its ability to image single cells in a high-throughput manner. However, there remains a challenge posed by the fundamental trade-off between throughput, sensitivity, and spatial resolution. Here we present deep-learning-enhanced imaging flow cytometry (dIFC) that circumvents this trade-off by implementing an image restoration algorithm on a virtual-freezing fluorescence imaging (VIFFI) flow cytometry platform, enabling higher throughput without sacrificing sensitivity and spatial resolution. A key component of dIFC is a high-resolution (HR) image generator that synthesizes "virtual" HR images from the corresponding low-resolution (LR) images acquired with a low-magnification lens (10×/0.4-NA). For IFC, a low-magnification lens is favorable because of reduced image blur of cells flowing at a higher speed, which allows higher throughput. We trained and developed the HR image generator with an architecture containing two generative adversarial networks (GANs). Furthermore, we developed dIFC as a method by combining the trained generator and IFC. We characterized dIFC using Chlamydomonas reinhardtii cell images, fluorescence in situ hybridization (FISH) images of Jurkat cells, and Saccharomyces cerevisiae (budding yeast) cell images, showing high similarities of dIFC images to images obtained with a high-magnification lens (40×/0.95-NA), at a high flow speed of 2 m s-1. We lastly employed dIFC to show enhancements in the accuracy of FISH-spot counting and neck-width measurement of budding yeast cells. These results pave the way for statistical analysis of cells with high-dimensional spatial information.

Label-free imaging flow cytometry for analysis and sorting of enzymatically dissociated tissues.

Biomedical research relies on identification and isolation of specific cell types using molecular biomarkers and sorting methods such as fluorescence or magnetic activated cell sorting. Labelling processes potentially alter the cells' properties and should be avoided, especially when purifying cells for clinical applications. A promising alternative is the label-free identification of cells based on physical properties. Sorting real-time deformability cytometry (soRT-DC) is a microfluidic technique for label-free analysis and sorting of single cells. In soRT-FDC, bright-field images of cells are analyzed by a deep neural net (DNN) to obtain a sorting decision, but sorting was so far only demonstrated for blood cells which show clear morphological differences and are naturally in suspension. Most cells, however, grow in tissues, requiring dissociation before cell sorting which is associated with challenges including changes in morphology, or presence of aggregates. Here, we introduce methods to improve robustness of analysis and sorting of single cells from nervous tissue and provide DNNs which can distinguish visually similar cells. We employ the DNN for image-based sorting to enrich photoreceptor cells from dissociated retina for transplantation into the mouse eye.

Machine learning assisted real-time deformability cytometry of CD34+ cells allows to identify patients with myelodysplastic syndromes.

Diagnosis of myelodysplastic syndrome (MDS) mainly relies on a manual assessment of the peripheral blood and bone marrow cell morphology. The WHO guidelines suggest a visual screening of 200 to 500 cells which inevitably turns the assessor blind to rare cell populations and leads to low reproducibility. Moreover, the human eye is not suited to detect shifts of cellular properties of entire populations. Hence, quantitative image analysis could improve the accuracy and reproducibility of MDS diagnosis. We used real-time deformability cytometry (RT-DC) to measure bone marrow biopsy samples of MDS patients and age-matched healthy individuals. RT-DC is a high-throughput (1000 cells/s) imaging flow cytometer capable of recording morphological and mechanical properties of single cells. Properties of single cells were quantified using automated image analysis, and machine learning was employed to discover morpho-mechanical patterns in thousands of individual cells that allow to distinguish healthy vs. MDS samples. We found that distribution properties of cell sizes differ between healthy and MDS, with MDS showing a narrower distribution of cell sizes. Furthermore, we found a strong correlation between the mechanical properties of cells and the number of disease-determining mutations, inaccessible with current diagnostic approaches. Hence, machine-learning assisted RT-DC could be a promising tool to automate sample analysis to assist experts during diagnosis or provide a scalable solution for MDS diagnosis to regions lacking sufficient medical experts.

AIDeveloper: Deep Learning Image Classification in Life Science and Beyond.

Artificial intelligence (AI)-based image analysis has increased drastically in recent years. However, all applications use individual solutions, highly specialized for a particular task. Here, an easy-to-use, adaptable, and open source software, called AIDeveloper (AID) to train neural nets (NN) for image classification without the need for programming is presented. AID provides a variety of NN-architectures, allowing to apply trained models on new data, obtain performance metrics, and export final models to different formats. AID is benchmarked on large image datasets (CIFAR-10 and Fashion-MNIST). Furthermore, models are trained to distinguish areas of differentiated stem cells in images of cell culture. A conventional blood cell count and a blood count obtained using an NN are compared, trained on >1.2 million images, and demonstrated how AID can be used for label-free classification of B- and T-cells. All models are generated by non-programmers on generic computers, allowing for an interdisciplinary use.

Real-Time Deformability Cytometry Detects Leukocyte Stiffening After Gadolinium-Based Contrast Agent Exposure.

OBJECTIVES: Reports on gadolinium (Gd) retention in soft tissues after administration of Gd-based contrast agents (GBCAs) raise concerns about Gd-induced changes in the biophysical properties of cells and tissues. Here, we investigate if clinical GBCAs of both classes of linear and macrocyclic structure cause changes in the mechanical properties of leukocytes in human blood samples. MATERIAL AND METHODS: Real-time deformability cytometry was applied to human blood samples from 6 donors. The samples were treated with 1 mM gadoteric acid (Dotarem), gadopentetic acid (Magnevist), gadobutrol (Gadovist), or Gd trichloride at 37°C for 1 hour to mimic clinical doses of GBCAs and exposure times. Leukocyte subtypes-lymphocytes, monocytes, and neutrophils-were identified based on their size and brightness and analyzed for deformability, which is inversely correlated with cellular stiffness. RESULTS: We observed significant stiffening (3%-13%, P < 0.01) of all investigated leukocyte subtypes, which was most pronounced for lymphocytes, followed by neutrophils and monocytes, and the effects were independent of the charge and steric structure of the GBCA applied. In contrast, no changes in cell size and brightness were observed, suggesting that deformability and cell stiffness measured by real-time deformability cytometry are sensitive to changes in the physical phenotypes of leukocytes after GBCA exposure. CONCLUSIONS: Real-time deformability cytometry might provide a quantitative blood marker for critical changes in the physical properties of blood cells in patients undergoing GBCA-enhanced magnetic resonance imaging.

Intelligent image-based deformation-assisted cell sorting with molecular specificity.

Although label-free cell sorting is desirable for providing pristine cells for further analysis or use, current approaches lack molecular specificity and speed. Here, we combine real-time fluorescence and deformability cytometry with sorting based on standing surface acoustic waves and transfer molecular specificity to image-based sorting using an efficient deep neural network. In addition to general performance, we demonstrate the utility of this method by sorting neutrophils from whole blood without labels.

Reliable isolation of human mesenchymal stromal cells from bone marrow biopsy specimens in patients after allogeneic hematopoietic cell transplantation.

Isolation of mesenchymal stromal cells (MSCs) from pretreated, hematologic patients is challenging. Especially after allogeneic hematopoietic cell transplantation (HCT), standard protocols using bone marrow aspirates fail to reliably recover sufficient cell numbers. Because MSCs are considered to contribute to processes that mainly affect the outcome after transplantation, such as an efficient lymphohematopoietic recovery, extent of graft-versus-host disease as well as the occurrence of leukemic relapse, it is of great clinical relevance to investigate MSC function in this context. Previous studies showed that MSCs can be isolated by collagenase digestion of large bone fragments of hematologically healthy patients undergoing hip replacement or knee surgeries. We have now further developed this procedure for the isolation of MSCs from hematologic patients after allogeneic HCT by using trephine biopsy specimens obtained during routine examinations. Comparison of aspirates and trephine biopsy specimens from patients after allogeneic HCT revealed a significantly higher frequency of clonogenic MSCs (colony-forming unit-fibroblast [CFU-F]) in trephine biopsy specimens (mean, 289.8 ± standard deviation 322.5 CFU-F colonies/1 × 106 total nucleated cells versus 4.2 ± 9.9; P < 0.0001). Subsequent expansion of functional MSCs isolated from trephine biopsy specimen was more robust and led to a significantly higher yield compared with control samples expanded from aspirates (median, 1.6 × 106; range, 0-2.3 × 107 P0 MSCs versus 5.4 × 104; range, 0-8.9 × 106; P < 0.0001). Using trephine biopsy specimens as MSC source facilitates the investigation of various clinical questions.

Targeting Mechanoresponsive Proteins in Pancreatic Cancer: 4-Hydroxyacetophenone Blocks Dissemination and Invasion by Activating MYH14.

Metastasis is complex, involving multiple genetic, epigenetic, biochemical, and physical changes in the cancer cell and its microenvironment. Cells with metastatic potential are often characterized by altered cellular contractility and deformability, lending them the flexibility to disseminate and navigate through different microenvironments. We demonstrate that mechanoresponsiveness is a hallmark of pancreatic cancer cells. Key mechanoresponsive proteins, those that accumulate in response to mechanical stress, specifically nonmuscle myosin IIA (MYH9) and IIC (MYH14), α-actinin 4, and filamin B, were highly expressed in pancreatic cancer as compared with healthy ductal epithelia. Their less responsive sister paralogs-myosin IIB (MYH10), α-actinin 1, and filamin A-had lower expression differential or disappeared with cancer progression. We demonstrate that proteins whose cellular contributions are often overlooked because of their low abundance can have profound impact on cell architecture, behavior, and mechanics. Here, the low abundant protein MYH14 promoted metastatic behavior and could be exploited with 4-hydroxyacetophenone (4-HAP), which increased MYH14 assembly, stiffening cells. As a result, 4-HAP decreased dissemination, induced cortical actin belts in spheroids, and slowed retrograde actin flow. 4-HAP also reduced liver metastases in human pancreatic cancer-bearing nude mice. Thus, increasing MYH14 assembly overwhelms the ability of cells to polarize and invade, suggesting targeting the mechanoresponsive proteins of the actin cytoskeleton as a new strategy to improve the survival of patients with pancreatic cancer. SIGNIFICANCE: This study demonstrates that mechanoresponsive proteins become upregulated with pancreatic cancer progression and that this system of proteins can be pharmacologically targeted to inhibit the metastatic potential of pancreatic cancer cells.

Analysis of Biomechanical Properties of Hematopoietic Stem and Progenitor Cells Using Real-Time Fluorescence and Deformability Cytometry.

Stem cell mechanics, determined predominantly by the cell's cytoskeleton, plays an important role in different biological processes such as stem cell differentiation or migration. Several methods to measure mechanical properties of cells are currently available, but most of them are limited in the ability to screen large heterogeneous populations in a robust and efficient manner-a feature required for successful translational applications. With real-time fluorescence and deformability cytometry (RT-FDC), mechanical properties of cells in suspension can be screened continuously at rates of up to 1,000 cells/s-similar to conventional flow cytometers-which makes it a suitable method not only for basic research but also for a clinical setting. In parallel to mechanical characterization, RT-FDC allows to measure specific molecular markers using standard fluorescence labeling. In this chapter, we provide a detailed protocol for the characterization of hematopoietic stem and progenitor cells (HSPCs) in heterogeneous mobilized peripheral blood using RT-FDC and present a specific morpho-rheological fingerprint of HSPCs that allows to distinguish them from all other blood cell types.

CASP1 variants influence subcellular caspase-1 localization, pyroptosome formation, pro-inflammatory cell death and macrophage deformability.

CASP1 variants result in reduced enzymatic activity of procaspase-1 and impaired IL-1ß release. Despite this, affected individuals can develop systemic autoinflammatory disease. These seemingly contradictory observations have only partially been explained by increased NF-κB activation through prolonged interaction of variant procaspase-1 with RIP2. To identify further disease underlying pathomechanisms, we established an in vitro model using shRNA-directed knock-down of procaspase-1 followed by viral transduction of human monocytes (THP-1) with plasmids encoding for wild-type procaspase-1, disease-associated CASP1 variants (p.L265S, p.R240Q) or a missense mutation in the active center of procaspase-1 (p.C285A). THP1-derived macrophages carrying CASP1 variants exhibited mutation-specific molecular alterations. We here provide in vitro evidence for abnormal pyroptosome formation (p.C285A, p.240Q, p.L265S), impaired nuclear (pro)caspase-1 localization (p.L265S), reduced pro-inflammatory cell death (p.C285A) and changes in macrophage deformability that may contribute to disease pathophysiology of patients with CASP1 variants. This offers previously unknown molecular pathomechanisms in patients with systemic autoinflammatory disease.