Adaptive Knowledge Distillation for High-Quality Unsupervised MRI Reconstruction with Model-Driven Priors.

Magnetic Resonance Imaging (MRI) reconstruction has made significant progress with the introduction of Deep Learning (DL) technology combined with Compressed Sensing (CS). However, most existing methods require large fully sampled training datasets to supervise the training process, which may be unavailable in many applications. Current unsupervised models also show limitations in performance or speed and may face unaligned distributions during testing. This paper proposes an unsupervised method to train competitive reconstruction models that can generate high-quality samples in an end-to-end style. Firstly teacher models are trained by filling the re-undersampled images and compared with the undersampled images in a self-supervised manner. The teacher models are then distilled to train another cascade model that can leverage the entire undersampled k-space during its training and testing. Additionally, we propose an adaptive distillation method to re-weight the samples based on the variance of teachers, which represents the confidence of the reconstruction results, to improve the quality of distillation. Experimental results on multiple datasets demonstrate that our method significantly accelerates the inference process while preserving or even improving the performance compared to the teacher model. In our tests, the distilled models show 5%-10% improvements in PSNR and SSIM compared with no distillation and are 10 times faster than the teacher. Source code of our work could be acquired at https://github.com/BITwzl/unsupervised_mri_reconstruction.

Resveratrol induces proliferation and differentiation of mouse pre-osteoblast MC3T3-E1 by promoting autophagy.

BACKGROUND: In mouse, it was discovered that resveratrol (Res) enhanced osteoporosis (OP) by boosting osteogenesis. Besides, Res can also have an impact on MC3T3-E1 cells, which are crucial for the control of osteogenesis and thus increase osteogenesis. Although some articles have discovered that Res enhanced autophagy to promote the value-added differentiation of MC3T3, it is unclear exactly how this affects the process of osteogenesis in mouse. Therefore, we will show that Res encourages MC3T3-E1 proliferation and differentiation in mouse pre-osteoblasts and further investigate the autophagy-related mechanism for this impact. METHODS: (1) MC3T3-E1 cells were separated into blank control group and various concentrations (0.01, 0.1, 1, 10, 100µmol/L) of group in order to determine the ideal Res concentration. In the Res group, Cell Counting Kit-8 (CCK-8) was used to measure the proliferation activity of pre-osteoblasts in mice in each group after resveratrol intervention. Alkaline Phosphatase (ALP) and alizarin red staining were used to gauge the degree of osteogenic differentiation, and RT-qPCR was used to measure the expression levels of Runx2 and OCN in the osteogenic differentiation ability of the cells. (2) In the experiment, four groups were set up: the control group, 3MA group, Res group, and Res + 3MA group. To examine cell mineralization, ALP and alizarin red staining were utilized. RT-qPCR and Western blot detection of cell autophagy activity levels and osteogenic differentiation capacity in each group following intervention. RESULTS: (1) Resveratrol might increase the number of mice pre-osteoblast, with the impact being most pronounced at 10µmol/L (P < 0.05). The nodules developed substantially more often than in the blank control group, and Runx2 and OCN expressions significantly increased (P < 0.05). (2) In contrast to the Res group, after 3MA purine blocked autophagy, the Res + 3MA group's alkaline phosphatase staining and the development of mineralized nodules were reduced. Runx2, OCN, LC3II / LC3I expression decreased, p62 expression increased (P < 0.05). CONCLUSION: The present study partially or indirectly demonstrated that Res may, through increased autophagy, induce osteogenic differentiation of MC3T3-E1 cells.

Deep learning based MRI reconstruction with transformer.

Magnetic resonance imaging (MRI) has become one of the most powerful imaging techniques in medical diagnosis, yet the prolonged scanning time becomes a bottleneck for application. Reconstruction methods based on compress sensing (CS) have made progress in reducing this cost by acquiring fewer points in the k-space. Traditional CS methods impose restrictions from different sparse domains to regularize the optimization that always requires balancing time with accuracy. Neural network techniques enable learning a better prior from sample pairs and generating the results in an analytic way. In this paper, we propose a deep learning based reconstruction method to restore high-quality MRI images from undersampled k-space data in an end-to-end style. Unlike prior literature adopting convolutional neural networks (CNN), advanced Swin Transformer is used as the backbone of our work, which proved to be powerful in extracting deep features of the image. In addition, we combined the k-space consistency in the output and further improved the quality. We compared our models with several reconstruction methods and variants, and the experiment results proved that our model achieves the best results in samples at low sampling rates. The source code of KTMR could be acquired at https://github.com/BITwzl/KTMR.

Endoglycosidase assay using enzymatically synthesized fluorophore-labeled glycans as substrates to uncover enzyme substrate specificities.

Glycan synthesis and degradation are not template but enzyme only driven processes. Substrate specificities of glyco-enzymes determine the structures of specific natural glycans. Using endoglycosidases as examples, we describe methods to study these enzymes. Endoglycosidase S/S2 specifically deglycosylates the conserved N-glycans of human immunoglobulin G. Endo-ß-Galactosidase hydrolyzes internal ß-galactosyl linkage in polylactosaminoglycan structures. To assay these enzymes, eleven fluorophore-labeled N-glycans and one polylactosamine ladder are synthesized. Digestion of these glycans result in mobility shift in gel electrophoresis. Results on Endo S/S2 assays reveal that they are most active on the agalactosylated biantennary N-glycans with decreased activity on galactosylated and sialylated glycans and little or no activity on branched and bisected glycans. Assays on Endo-ß-Gal reveal that the enzyme is active from pH 3.5 to 9.0 and the ß3-linked GlcNAc adjacent to the cleavage site is minimal for the enzyme recognition with the optimal recognition motif spanning at least four lactosamine repeats. Our methods will provide an opportunity to understand how specific glycans are synthesized and degraded.

Fluorescent glycan fingerprinting of SARS2 spike proteins.

Glycosylation is the most common post-translational modification and has myriad of biological functions. However, glycan analysis has always been a challenge. Here, we would like to present new techniques for glycan fingerprinting based on enzymatic fluorescent labeling and gel electrophoresis. The method is illustrated on SARS2 spike (S) glycoproteins. SARS2, a novel coronavirus and the causative agent of the COVID-19 pandemic, has had significant social and economic impacts since the end of 2019. To obtain the N-glycan fingerprint of an S protein, glycans released from the protein are first labeled through enzymatic incorporation of fluorophore-conjugated sialic acid or fucose, then separated by SDS-PAGE, and finally visualized with a fluorescent imager. To identify the labeled glycans of a fingerprint, glycan standards and glycan ladders are enzymatically generated and run alongside the samples as references. By comparing the mobility of a labeled glycan to that of a glycan standard, the identity of glycans maybe determined. O-glycans can also be fingerprinted. Due to the lack of an enzyme for broad O-glycan release, O-glycans on the S protein can be labeled with fluorescent sialic acid and digested with trypsin to obtain labeled glycan peptides that are then separated by gel electrophoresis. Glycan fingerprinting could serve as a quick method for globally assessing the glycosylation of a specific glycoprotein.

Assays for hyaluronidases and heparanase using nonreducing end fluorophore-labeled hyaluronan and heparan sulfate proteoglycan.

Glycosaminoglycans (GAGs), such as hyaluronan (HA) and heparan sulfate (HS), are a large group of polysaccharides found in the extracellular matrix and on the cell surface. The turnover of these molecules is controlled by de novo synthesis and catabolism through specific endoglycosidases, which are the keys to our understanding of the homeostasis of GAGs and could hold opportunities for therapeutic intervention. Herein, we describe assays for endoglycosidases using nonreducing end fluorophore-labeled GAGs, in which GAGs were labeled via incorporation of GlcNAz by specific synthases and cycloaddition of alkyne fluorophores and then digested with corresponding endoglycosidases. Assays of various HA-specific hyaluronidases (HYALs), including PH-20 or SPAM1, and HS-specific heparanase (HPSE) are presented. We demonstrated the distinctive pH profiles, substrate specificities and specific activities of these enzymes and provided evidence that both HYAL3 and HYAL4 are authentic hyaluronidases. In addition, while all HYALs are active on high-molecular-weight HA, they are active on low-molecular-weight HA. Subsequently, we defined a new way of measuring the activities of HYALs. Our results indicate that the activities of HYALs must be under strict pH regulation. Our quantitative methods of measuring the activity GAG endoglycosidases could bring the opportunity of designing novel therapeutics by targeting these important enzymes.

Fluorescent Detection of O-GlcNAc via Tandem Glycan Labeling.

O-GlcNAcylation is a reversible serine/threonine glycosylation on cytosolic and nuclear proteins that are involved in various regulatory pathways. However, the detection and quantification of O-GlcNAcylation substrates have been challenging. Here, we report a highly efficient method for the identification of O-GlcNAc modification via tandem glycan labeling, in which O-GlcNAc is first galactosylated and then sialylated with a fluorophore-conjugated sialic acid residue, therefore enabling highly sensitive fluorescent detection. The method is validated on various proteins that are known to be modified by O-GlcNAcylation including CK2, NOD2, SREBP1c, AKT1, PKM, and PFKFB3, and on the nuclear extract of HEK293 cells. Using this method, we then report the evidence that hypoxia-inducible factor HIF1α is a potential target for O-GlcNAcylation, suggesting a possibly direct connection between the metabolic O-GlcNAc pathway and the hypoxia pathway.

Detecting substrate glycans of fucosyltransferases with fluorophore-conjugated fucose and methods for glycan electrophoresis.

Like sialylation, fucose usually locates at the nonreducing ends of various glycans on glycoproteins and constitutes important glycan epitopes. Detecting the substrate glycans of fucosyltransferases is important for understanding how these glycan epitopes are regulated in response to different growth conditions and external stimuli. Here we report the detection of these glycans on glycoproteins as well as in their free forms via enzymatic incorporation of fluorophore-conjugated fucose using FUT2, FUT6, FUT7, FUT8 and FUT9. Specifically, we describe the detection of the substrate glycans of these enzymes on fetal bovine fetuin, recombinant H1N1 viral neuraminidase and therapeutic antibodies. The detected glycans include complex and high-mannose N-glycans. By establishing a series of precursors for the synthesis of Lewis X and sialyl Lewis X structures, we not only provide convenient electrophoresis methods for studying glycosylation but also demonstrate the substrate specificities and some kinetic features of these enzymes. Our results support the notion that fucosyltransferases are key targets for regulating the synthesis of Lewis X and sialyl Lewis X structures.

Differential distribution of N- and O-Glycans and variable expression of sialyl-T antigen on HeLa cells-Revealed by direct fluorescent glycan imaging.

Cells are covered with glycans. The expression and distribution of specific glycans on the surface of a cell are important for various cellular functions. Imaging these glycans is essential to aid elucidation of their biological roles. Here, utilizing methods of direct fluorescent glycan imaging, in which fluorescent sialic acids are directly incorporated into substrate glycans via recombinant sialyltranferases, we report the differential distribution of N- and O-glycans and variable expression of sialyl-T antigen on HeLa cells. While the expression of N-glycans tends to be more peripheral at positions where cell-cell interaction occurs, O-glycan expression is more granular but relatively evenly distributed on positive cells. While N-glycans are expressed on all cells, sialyl-T antigen expression exhibits a wide spectrum of variation with some cells being strongly positive and some cells being almost completely negative. The differential distribution of N- and O-glycans on cell surface reflects their distinctive roles in cell biology.

Direct fluorescent glycan labeling with recombinant sialyltransferases.

Glycosylation is a common modification found on numerous proteins and lipids. However, direct detection of glycans on these intact biomolecules has been challenge. Here, utilizing enzymatic incorporation of fluorophore-conjugated sialic acids, dubbed as direct fluorescent glycan labeling, we report the labeling and detection of N- and O-glycans on glycoproteins. The method allows detection of specific glycans without the laborious gel blotting and chemiluminescence reactions used in Western blotting. The method can also be used with a variety of fluorescent dyes.

Detecting and Imaging O-GlcNAc Sites Using Glycosyltransferases: A Systematic Approach to Study O-GlcNAc.

O-GlcNAcylation is a reversible serine/threonine glycosylation for regulating protein activity and availability inside cells. In a given protein, O-GlcNAcylated and unoccupied O-linked ß-N-acetylglucosamine (O-GlcNAc) sites are referred to as closed and open sites, respectively. The balance between open and closed sites is believed to be dynamically regulated. In this report, closed sites are detected using in vitro incorporation of GalNAz by B3GALNT2, and open sites are detected by in vitro incorporation of GlcNAz by O-GlcNAc transferase (OGT), via click chemistry. For assessing total O-GlcNAc sites, a sample is O-GlcNAcylated in vitro by OGT before detecting by B3GALNT2. The methods are demonstrated on purified recombinant proteins including CK2, AKT1, and PFKFB3, and cellular extracts of HEK cells. Through O-GlcNAc imaging, the modification degree of O-GlcNAc in nuclei of Chinese hamster ovary cells was estimated. The detection and imaging of both open and closed O-GlcNAc sites provide a systematic approach to study this important post-translational modification.

Imaging specific cellular glycan structures using glycosyltransferases via click chemistry.

Heparan sulfate (HS) is a polysaccharide fundamentally important for biologically activities. T/Tn antigens are universal carbohydrate cancer markers. Here, we report the specific imaging of these carbohydrates using a mesenchymal stem cell line and human umbilical vein endothelial cells (HUVEC). The staining specificities were demonstrated by comparing imaging of different glycans and validated by either removal of target glycans, which results in loss of signal, or installation of target glycans, which results in gain of signal. As controls, representative key glycans including O-GlcNAc, lactosaminyl glycans and hyaluronan were also imaged. HS staining revealed novel architectural features of the extracellular matrix (ECM) of HUVEC cells. Results from T/Tn antigen staining suggest that O-GalNAcylation is a rate-limiting step for O-glycan synthesis. Overall, these highly specific approaches for HS and T/Tn antigen imaging should greatly facilitate the detection and functional characterization of these biologically important glycans.

Non-reducing end labeling of heparan sulfate via click chemistry and a high throughput ELISA assay for heparanase.

Heparan sulfate (HS) is a linear polysaccharide found in the extracellular matrix (ECM) and on the cell membrane. It plays numerous roles in cellular events, including cell growth, migration and differentiation through binding to various growth factors, cytokines and other ECM proteins. Heparanase (HPSE) is an endoglycosidase that cleaves HS in the ECM and cell membrane. By degrading HS, HPSE not only alters the integrity of the ECM but also releases growth factors and angiogenic factors bound to HS chains, therefore, changes various cellular activities, including cell mobility that is critical for cancer metastasis. Accordingly, HPSE is an ideal drug target for cancer therapeutics. Here, we describe a method for non-reducing end labeling of HS via click chemistry (CC), and further use it in a novel HPSE assay. HS chains on a recombinant human syndecan-4 are first labeled at their non-reducing ends with GlcNAz using dimeric HS polymerase EXT1/EXT2. The labeled sample is then biotinylated through CC, immobilized on a multi-well plate and detected with ELISA. HPSE digestion of the biotinylated sample removes the label and, therefore, reduces the signal in ELISA assay. Non-reducing end labeling avoids the interference in an HPSE reaction caused by any internal labeling of HS. The assay is very sensitive with only 2.5 ng of labeled syndecan-4 needed in each reaction. The assay is also highly reproducible with a Z' > 0.6. Overall, this new method is suitable for high-throughput drug screening on HPSE.

Glycoengineering of E-Selectin Ligands by Intracellular versus Extracellular Fucosylation Differentially Affects Osteotropism of Human Mesenchymal Stem Cells.

Human mesenchymal stem cells (MSCs) hold great promise in cellular therapeutics for skeletal diseases but lack expression of E-selectin ligands that direct homing of blood-borne cells to bone marrow. Previously, we described a method to engineer E-selectin ligands on the MSC surface by exofucosylating cells with fucosyltransferase VI (FTVI) and its donor sugar, GDP-Fucose, enforcing transient surface expression of the potent E-selectin ligand HCELL with resultant enhanced osteotropism of intravenously administered cells. Here, we sought to determine whether E-selectin ligands created via FTVI-exofucosylation are distinct in identity and function to those created by FTVI expressed intracellularly. To this end, we introduced synthetic modified mRNA encoding FTVI (FUT6-modRNA) into human MSCs. FTVI-exofucosylation (i.e., extracellular fucosylation) and FUT6-modRNA transfection (i.e., intracellular fucosylation) produced similar peak increases in cell surface E-selectin ligand levels, and shear-based functional assays showed comparable increases in tethering/rolling on human endothelial cells expressing E-selectin. However, biochemical analyses revealed that intracellular fucosylation induced expression of both intracellular and cell surface E-selectin ligands and also induced a more sustained expression of E-selectin ligands compared to extracellular fucosylation. Notably, live imaging studies to assess homing of human MSC to mouse calvarium revealed more osteotropism following intravenous administration of intracellularly-fucosylated cells compared to extracellularly-fucosylated cells. This study represents the first direct analysis of E-selectin ligand expression programmed on human MSCs by FTVI-mediated intracellular versus extracellular fucosylation. The observed differential biologic effects of FTVI activity in these two contexts may yield new strategies for improving the efficacy of human MSCs in clinical applications. Stem Cells 2016;34:2501-2511.

Core-6 fucose and the oligomerization of the 1918 pandemic influenza viral neuraminidase.

The 1918 H1N1 influenza virus was responsible for one of the most deadly pandemics in human history. Yet to date, the structure component responsible for its virulence is still a mystery. In order to search for such a component, the neuraminidase (NA) antigen of the virus was expressed, which led to the discovery of an active form (tetramer) and an inactive form (dimer and monomer) of the protein due to different glycosylation. In this report, the N-glycans from both forms were released and characterized by mass spectrometry. It was found that the glycans from the active form had 26% core-6 fucosylated, while the glycans from the inactive form had 82% core-6 fucosylated. Even more surprisingly, the stalk region of the active form was almost completely devoid of core-6-linked fucose. These findings were further supported by the results obtained from in vitro incorporation of azido fucose and (3)H-labeled fucose using core-6 fucosyltransferase, FUT8. In addition, the incorporation of fucose did not change the enzymatic activity of the active form, implying that core-6 fucose is not directly involved in the enzymatic activity. It is postulated that core-6 fucose prohibits the oligomerization and subsequent activation of the enzyme.

Probing sialoglycans on fetal bovine fetuin with azido-sugars using glycosyltransferases.

Sialic acids are negatively charged sugar residues commonly found on the terminal positions of most glycoproteins. They play important roles in the stability and solubility of these proteins. Due to their unique positioning, they also frequently act as receptors for various ligands, and therefore are involved in numerous cell-cell and cell-pathogen interactions. Here, using in vitro incorporation of clickable monosaccharides with various glycosyltransferases, we probed the sialoglycans on fetal bovine fetuin. The incorporated monosaccharides were detected with chemiluminescence via click chemistry in a format of western blotting. The results indicate that the non-reducing end Gal residues on both N- and O-glycans are fully sialylated, but the peptide-linked GalNAc residues in O-glycans are not. The presence of sialyl core-1 glycan was repeatedly confirmed by probing with α-2,3-sialyltransferases, N-acetylgalactosaminide α-2,6-sialyltransferases and a ß-1,6-N-acetylglucosaminyltransferase that is specific for core-1 glycan. The results also suggest the presence of a minute amount of sialyl Tn antigen on the protein.

Glycoprotein labeling with click chemistry (GLCC) and carbohydrate detection.

Molecular labeling and detection techniques are essential to research in life science. Here, a method for glycoprotein labeling/carbohydrate detection through glycan replacement, termed glycoprotein labeling with click chemistry (GLCC), is described. In this method, a glycoprotein is first treated with specific glycosidases to remove certain sugar residues, a procedure that creates acceptor sites for a specific glycosyltransferase. A 'clickable' monosaccharide is then installed onto these sites by the glycosyltransferase. This modified glycoprotein is then conjugated to a reporter molecule using a click chemistry reaction. For glycoproteins that already contain vacant glycosylation sites, deglycosylation is not needed before the labeling step. As a demonstration, labeling on fetal bovine fetuin, mouse immunoglobulin IgG and bacterial expressed human TNFα and TNFß are shown. Compared to traditional ways of protein labeling, labeling at glycosylation sites with GLCC is considerably more specific and less likely to have adverse effects, and, when utilized as a method for carbohydrate detection, this method is also highly specific and sensitive.

Nonradioactive glycosyltransferase and sulfotransferase assay to study glycosaminoglycan biosynthesis.

Glycosaminoglycans (GAGs) are linear polysaccharides with repeating disaccharide units. GAGs include heparin, heparan sulfate, chondroitin sulfate, dermatan sulfate, keratan sulfate, and hyaluronan. All GAGs, except for hyaluronan, are usually sulfated. GAGs are polymerized by mono- or dual-specific glycosyltransferases and sulfated by various sulfotransferases. To further our understanding of GAG chain length regulation and synthesis of specific sulfation motifs on GAG chains, it is imperative to understand the kinetics of GAG synthetic enzymes. Here, nonradioactive colorimetric enzymatic assays are described for these glycosyltransferases and sulfotransferases. In both cases, the leaving nucleotides or nucleosides are hydrolyzed using specific phosphatases, and the released phosphate is subsequently detected using malachite reagents.

A liquid chromatography-mass spectrometry-based approach to characterize the substrate specificity of mammalian heparanase.

Extracellular heparanase activity releases growth factors and angiogenic factors from heparan sulfate (HS) storage sites and alters the integrity of the extracellular matrix. These activities lead to a loss of normal cell matrix adherent junctions and correlate with invasive cellular phenotypes. Elevated expression of heparanase is associated with several human cancers and with vascular remodeling. Heparanase cleaves only a limited fraction of glucuronidic linkages in HS. There have been few investigations of the functional consequences of heparanase activity, largely due to the heterogeneity and complexity of HS. Here, we report a liquid chromatography-mass spectrometry (LC-MS)-based approach to profile the terminal structures created by heparanase digestion and reconstruct the heparanase cleavage sites from the products. Using this method, we demonstrate that heparanase cleaves at the non-reducing side of highly sulfated HS domains, exposing cryptic growth factor binding sites. This cleavage pattern is observed in HS from several tissue sources, regardless of overall sulfation degree, indicating a common recognition pattern. We further demonstrate that heparanase cleavage of HS chains leads to increased ability to support FGF2-dependent cell proliferation. These results suggest a new mechanism to explain how heparanase might potentiate the uncontrolled cell proliferation associated with cancer through its ability to activate nascent growth factor-promoting domains within HS.

Detecting O-GlcNAc using in vitro sulfation.

O-linked ß-N-acetylglucosamine (O-GlcNAc) glycosylation, the covalent attachment of N-acetylglucosamine to serine and threonine residues of proteins, is a post-translational modification that shares many features with protein phosphorylation. O-GlcNAc is essential for cell survival and plays important role in many biological processes (e.g. transcription, translation, cell division) and human diseases (e.g. diabetes, Alzheimer's disease, cancer). However, detection of O-GlcNAc is challenging. Here, a method for O-GlcNAc detection using in vitro sulfation with two N-acetylglucosamine (GlcNAc)-specific sulfotransferases, carbohydrate sulfotransferase 2 and carbohydrate sulfotransferase 4, and the radioisotope (35)S is described. Sulfation on free GlcNAc is first demonstrated, and then on O-GlcNAc residues of peptides as well as nuclear and cytoplasmic proteins. It is also demonstrated that the sulfation on O-GlcNAc is sensitive to OGT and O-ß-N-acetylglucosaminidase treatment. The labeled samples are separated on sodium dodecyl sulfate-polyacrylamide gel electrophoresis and visualized by autoradiography. Overall, the method is sensitive, specific and convenient.