Digital Staining of White Blood Cells With Confidence Estimation.

Chemical staining of the blood smears is one of the crucial components of blood analysis. It is an expensive, lengthy and sensitive process, often prone to produce slight variations in colour and seen structures due to a lack of unified protocols across laboratories. Even though the current developments in deep generative modeling offer an opportunity to replace the chemical process with a digital one, there are specific safety-ensuring requirements due to the severe consequences of mistakes in a medical setting. Therefore digital staining system would profit from an additional confidence estimation quantifying the quality of the digitally stained white blood cell. To this aim, during the staining generation, we disentangle the latent space of the Generative Adversarial Network, obtaining separate representation s of the white blood cell and the staining. We estimate the generated image's confidence of white blood cell structure and staining quality by corrupting these representations with noise and quantifying the information retained between multiple outputs. We show that confidence estimated in this way correlates with image quality measured in terms of LPIPS values calculated for the generated and ground truth stained images. We validate our method by performing digital staining of images captured with a Differential Inference Contrast microscope on a dataset composed of white blood cells of 24 patients. The high absolute value of the correlation between our confidence score and LPIPS demonstrates the effectiveness of our method, opening the possibility of predicting the quality of generated output and ensuring trustworthiness in medical safety-critical setup.

Multi-Task Multi-Domain Learning for Digital Staining and Classification of Leukocytes.

This paper addresses digital staining and classification of the unstained white blood cell images obtained with a differential contrast microscope. We have data coming from multiple domains that are partially labeled and partially matching across the domains. Using unstained images removes time-consuming staining procedures and could facilitate and automatize comprehensive diagnostics. To this aim, we propose a method that translates unstained images to realistically looking stained images preserving the inter-cellular structures, crucial for the medical experts to perform classification. We achieve better structure preservation by adding auxiliary tasks of segmentation and direct reconstruction. Segmentation enforces that the network learns to generate correct nucleus and cytoplasm shape, while direct reconstruction enforces reliable translation between the matching images across domains. Besides, we build a robust domain agnostic latent space by injecting the target domain label directly to the generator, i.e., bypassing the encoder. It allows the encoder to extract features independently of the target domain and enables an automated domain invariant classification of the white blood cells. We validated our method on a large dataset composed of leukocytes of 24 patients, achieving state-of-the-art performance on both digital staining and classification tasks.

Genome Wide Methylome Alterations in Lung Cancer.

Aberrant cytosine 5-methylation underlies many deregulated elements of cancer. Among paired non-small cell lung cancers (NSCLC), we sought to profile DNA 5-methyl-cytosine features which may underlie genome-wide deregulation. In one of the more dense interrogations of the methylome, we sampled 1.2 million CpG sites from twenty-four NSCLC tumor (T)-non-tumor (NT) pairs using a methylation-sensitive restriction enzyme- based HELP-microarray assay. We found 225,350 differentially methylated (DM) sites in adenocarcinomas versus adjacent non-tumor tissue that vary in frequency across genomic compartment, particularly notable in gene bodies (GB; p<2.2E-16). Further, when DM was coupled to differential transcriptome (DE) in the same samples, 37,056 differential loci in adenocarcinoma emerged. Approximately 90% of the DM-DE relationships were non-canonical; for example, promoter DM associated with DE in the same direction. Of the canonical changes noted, promoter (PR) DM loci with reciprocal changes in expression in adenocarcinomas included HBEGF, AGER, PTPRM, DPT, CST1, MELK; DM GB loci with concordant changes in expression included FOXM1, FERMT1, SLC7A5, and FAP genes. IPA analyses showed adenocarcinoma-specific promoter DMxDE overlay identified familiar lung cancer nodes [tP53, Akt] as well as less familiar nodes [HBEGF, NQO1, GRK5, VWF, HPGD, CDH5, CTNNAL1, PTPN13, DACH1, SMAD6, LAMA3, AR]. The unique findings from this study include the discovery of numerous candidate The unique findings from this study include the discovery of numerous candidate methylation sites in both PR and GB regions not previously identified in NSCLC, and many non-canonical relationships to gene expression. These DNA methylation features could potentially be developed as risk or diagnostic biomarkers, or as candidate targets for newer methylation locus-targeted preventive or therapeutic agents.

High-throughput library screening identifies two novel NQO1 inducers in human lung cells.

Many phytochemicals possess antioxidant and cancer-preventive properties, some putatively through antioxidant response element-mediated phase II metabolism, entailing mutagen/oxidant quenching. In our recent studies, however, most candidate phytochemical agents were not potent in inducing phase II genes in normal human lung cells. In this study, we applied a messenger RNA (mRNA)-specific gene expression-based high throughput in vitro screening approach to discover new, potent plant-derived phase II inducing chemopreventive agents. Primary normal human bronchial epithelial (NHBE) cells and immortalized human bronchial epithelial cells (HBECs) were exposed to 800 individual compounds in the MicroSource Natural Products Library. At a level achievable in humans by diet (1.0 µM), 2,3-dihydroxy-4-methoxy-4'-ethoxybenzophenone (DMEBP), triacetylresveratrol (TRES), ivermectin, sanguinarine sulfate, and daunorubicin induced reduced nicotinamide adenine dinucleotide phosphate:quinone oxidoreductase 1 (NQO1) mRNA and protein expression in NHBE cells. DMEBP and TRES were the most attractive agents as coupling potency and low toxicity for induction of NQO1 (mRNA level, ≥3- to 10.8-fold that of control; protein level, ≥ two- to fourfold that of control). Induction of glutathione S-transferase pi mRNA expression was modest, and none was apparent for glutathione S-transferase pi protein expression. Measurements of reactive oxygen species and glutathione/oxidized glutathione ratio showed an antioxidant effect for DMEBP, but no definite effect was found for TRES in NHBE cells. Exposure of NHBE cells to H(2)O(2) induced nuclear translocation of nuclear factor erythroid 2-related factor 2, but this translocation was not significantly inhibited by TRES and DMEBP. These studies show that potency and low toxicity may align for two potential NQO1-inducing agents, DMEBP and TRES.

Mapping the N-terminal residues of Epstein-Barr virus gp42 that bind gH/gL by using fluorescence polarization and cell-based fusion assays.

Epstein-Barr virus (EBV) requires at a minimum membrane-associated glycoproteins gB, gH, and gL for entry into host cells. B-cell entry additionally requires gp42, which binds to gH/gL and triggers viral entry into B cells. The presence of soluble gp42 inhibits membrane fusion with epithelial cells by forming a stable heterotrimer of gH/gL/gp42. The interaction of gp42 with gH/gL has been previously mapped to residues 36 to 81 at the N-terminal region of gp42. In this study, we further mapped this region to identify essential features for binding to gH/gL by use of synthetic peptides. Data from fluorescence polarization, cell-cell fusion, and viral infection assays demonstrated that 33 residues corresponding to 44 to 61 and 67 to 81 of gp42 were indispensable for maintaining low-nanomolar-concentration gH/gL binding affinity and inhibiting B-cell fusion and epithelial cell fusion as well as viral infection. Overall, specific, large hydrophobic side chain residues of gp42 appeared to provide critical interactions, determining the binding strength. Mutations of these residues also diminished the inhibition of B-cell and epithelial cell fusions as well as EBV infection. A linker region (residues 62 to 66) between two gH/gL binding regions served as an important spacer, but individual amino acids were not critical for gH/gL binding. Probing the binding site of gH/gL and gp42 with gp42 peptides is critical for a better understanding of the interaction of gH/gL with gp42 as well as for the design of novel entry inhibitors of EBV and related human herpesviruses.

The contribution of systems biology and reverse genetics to the understanding of Kaposi's sarcoma-associated herpesvirus pathogenesis in endothelial cells.

Kaposi's sarcoma-associated herpesvirus (KSHV)/human herpesvirus-8 is the causative agent of the endothelial cell-derived tumour Kaposi's sarcoma. Herpesviruses possess large complex genomes which provide many options to regulate cellular physiology during the viral life cycle and in the course of tumourigenicity. Novel techniques of systems biology and reverse genetics are increasingly applied to dissect the complex interaction of KSHV with endothelial cells. This review will outline novel results and pitfalls of these technologies in the elucidation of KSHV pathogenicity.

A systems biology approach to identify the combination effects of human herpesvirus 8 genes on NF-kappaB activation.

Human herpesvirus 8 (HHV-8) is the etiologic agent of Kaposi's sarcoma and primary effusion lymphoma. Activation of the cellular transcription factor nuclear factor-kappa B (NF-kappaB) is essential for latent persistence of HHV-8, survival of HHV-8-infected cells, and disease progression. We used reverse-transfected cell microarrays (RTCM) as an unbiased systems biology approach to systematically analyze the effects of HHV-8 genes on the NF-kappaB signaling pathway. All HHV-8 genes individually (n = 86) and, additionally, all K and latent genes in pairwise combinations (n = 231) were investigated. Statistical analyses of more than 14,000 transfections identified ORF75 as a novel and confirmed K13 as a known HHV-8 activator of NF-kappaB. K13 and ORF75 showed cooperative NF-kappaB activation. Small interfering RNA-mediated knockdown of ORF75 expression demonstrated that this gene contributes significantly to NF-kappaB activation in HHV-8-infected cells. Furthermore, our approach confirmed K10.5 as an NF-kappaB inhibitor and newly identified K1 as an inhibitor of both K13- and ORF75-mediated NF-kappaB activation. All results obtained with RTCM were confirmed with classical transfection experiments. Our work describes the first successful application of RTCM for the systematic analysis of pathofunctions of genes of an infectious agent. With this approach, ORF75 and K1 were identified as novel HHV-8 regulatory molecules on the NF-kappaB signal transduction pathway. The genes identified may be involved in fine-tuning of the balance between latency and lytic replication, since this depends critically on the state of NF-kappaB activity.

Viral inhibitor of apoptosis vFLIP/K13 protects endothelial cells against superoxide-induced cell death.

Human herpesvirus 8 (HHV-8) is the etiological agent of Kaposi's sarcoma (KS). HHV-8 encodes an antiapoptotic viral Fas-associated death domain-like interleukin-1beta-converting enzyme-inhibitory protein (vFLIP/K13). The antiapoptotic activity of vFLIP/K13 has been attributed to an inhibition of caspase 8 activation and more recently to its capability to induce the expression of antiapoptotic proteins via activation of NF-kappaB. Our study provides the first proteome-wide analysis of the effect of vFLIP/K13 on cellular-protein expression. Using comparative proteome analysis, we identified manganese superoxide dismutase (MnSOD), a mitochondrial antioxidant and an important antiapoptotic enzyme, as the protein most strongly upregulated by vFLIP/K13 in endothelial cells. MnSOD expression was also upregulated in endothelial cells upon infection with HHV-8. Microarray analysis confirmed that MnSOD is also upregulated at the RNA level, though the differential expression at the RNA level was much lower (5.6-fold) than at the protein level (25.1-fold). The induction of MnSOD expression was dependent on vFLIP/K13-mediated activation of NF-kappaB, occurred in a cell-intrinsic manner, and was correlated with decreased intracellular superoxide accumulation and increased resistance of endothelial cells to superoxide-induced death. The upregulation of MnSOD expression by vFLIP/K13 may support the survival of HHV-8-infected cells in the inflammatory microenvironment in KS.