Journal: Journal of Experimental & Clinical Cancer Research : CR

Article Title: The SF3B1 R625H mutation promotes prolactinoma tumor progression through aberrant splicing of DLG1

doi: 10.1186/s13046-022-02245-0

Figure Lengend Snippet: Mutant SF3B1 promoted EMT phenotypes in prolactinoma. A GO functional enrichment analysis of DEGs in Sf3b1 mutant vs. wild-type GH3 cells. B Scanning electron microscopy (SEM; upper) and confocal images of the actin cytoskeleton stained with rhodamine–phalloidin (lower) of Sf3b1 wild-type and mutant GH3 cells; scale bar: 10 μm. C Scratch wound healing assays were performed using Sf3b1 wild-type and mutant GH3 cells; scale bar: 200 μm. Relative scratch widths are shown over time ( n = 3; mean ± SD shown below; P values by Student’s test). D Kinetics of wound confluence over 96 h were analyzed using IncuCyte 2018B software (Essen Bioscience). The results are expressed as mean ± SD, n = 3. P values by two-way ANOVA; **** P < 0.0001. E DLG1, E-cadherin, N-cadherin, Vimentin, and Snail levels in Sf3b1 wild-type and mutant GH3 cells transduced with or without Dlg1-FLAG cDNA. G Immunofluorescence staining of E-cadherin and Vimentin in rat prolactinoma tumors stereotactically injected with Ad-null, Ad-SF3B1 WT , and Ad-SF3B1 R625H ; scale bar: 50 μm. F Representative immunohistochemical staining for E-cadherin, N-cadherin, and Snail in SF3B1 mutant and wild-type human prolactinoma tumors; scale bar: 50 μm. H Venn diagram of the numbers of differentially spliced (q < 0.05; t test) and differentially expressed genes (q < 0.05; DESeq2) in SF3B1 mutant human prolactinoma and GH3 samples

Article Snippet: Phase contrast imaging was performed every 12 h till 96 h. Images were analyzed using IncuCyte® S3 2018B-2019A software (Essen Bioscience), and data were analyzed using GraphPad Prism7 (GraphPad Software, Inc., La Jolla, CA, USA).

Techniques: Mutagenesis, Functional Assay, Electron Microscopy, Staining, Software, Transduction, Immunofluorescence, Injection, Immunohistochemistry

Journal: Scientific Reports

Article Title: Investigating heterogeneities of live mesenchymal stromal cells using AI-based label-free imaging

doi: 10.1038/s41598-021-85905-z

Figure Lengend Snippet: Convolutional neural network (CNN) training for predicting marker expression. ( a ) Schematic of an iterative machine learning training operation. The generator produces a virtually stained prediction image using a U-Net architecture. The discriminator measures the probability of similarity between prediction and target and updates loss function parameters. ( b ) Left to right: example of a phase contrast input image, a target (ground truth) immunofluorescence image and the respective prediction image. ( c ) Individually trained marker predictions from a single transmitted light microscopy image combined into a multi-marker composite. ( d , e ) Performance of virtual surface marker prediction using the U-Net + cGAN model (orange) versus the U-Net only model (grey). Both ( d ), the Source Pearson correlation coefficient \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{s}$$\end{document} r s , and ( e ), the Laplacian Pearson correlation coefficient \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{lap}$$\end{document} r lap , demonstrate higher values over all channels. Each data point represents one target-prediction image pair, with outliers shown as diamonds. Boxplots were created by using OriginLab version 2019.

Article Snippet: Subfigures were generated using the OriginLab software version 2019 ( a – c ) and Fiji ImageJ software ( d ).

Techniques: Marker, Expressing, Staining, Immunofluorescence, Light Microscopy

Journal: Scientific Reports

Article Title: Investigating heterogeneities of live mesenchymal stromal cells using AI-based label-free imaging

doi: 10.1038/s41598-021-85905-z

Figure Lengend Snippet: Characterization and robustness analysis of the proposed ML model with CD105 staining. ( a ) Comparison of a target image with prediction images resulting from a model trained with 20 or 1280 images. The ML training set consisting of 20 images accurately predicts boundaries of cells but fails to identify small spindle-like subcellular structures (arrows). ( b , c ) Performance of fluorescence image predictions measured by image-wise Source and Laplacian Pearson correlation coefficients as a function of the number of training images. A higher number of training images improve the predictions of details and rare events. For both boxplots, ns stands for not significant; *p < 0.05; **p < 0.001; ***p < 0.0001. ( d ) Simulating weakly expressed markers with step-wise lowering illumination. The first column displays immunofluorescent target images and the second and third columns represent a magnified target and prediction image, respectively, with adjusted brightness for better visualization. ( e , f ) Prediction performance across different excitation levels. Dim markers can be represented fairly robustly by our AI algorithm, with prediction differences primarily visible for small scale features such as detailed protein localization and fine morphological shapes. Dashed line in ( e ) indicates the theoretical maximum correlation between the target and optimal prediction images ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C_{max}$$\end{document} C max ). Error bars represent the standard deviation of the mean. All box and point plots were created by using OriginLab version 2019.

Article Snippet: Subfigures were generated using the OriginLab software version 2019 ( a – c ) and Fiji ImageJ software ( d ).

Techniques: Staining, Fluorescence, Standard Deviation

Journal: Scientific Reports

Article Title: Investigating heterogeneities of live mesenchymal stromal cells using AI-based label-free imaging

doi: 10.1038/s41598-021-85905-z

Figure Lengend Snippet: Multi-marker heterogeneity and gene expression-morphology correlation. ( a ) Example of cell outlining and marker distribution for single cell analysis. ( b ) Histogram of standardized intensity distribution of 500 single cells for 8 MSC markers. Percentages illustrate the standard deviation of each intensity distribution. Histogram was created by using OriginLab version 2019. ( c ) Hierarchical clustering of marker intensities and 12 different morphological features analyzed from outlined cells. Marker intensity and morphological features lead to distinct clustering patterns. The color represents the fold change with respect to the mean value for each row. Clustering analysis (heatmap) was performed using Python version 3.8 (seaborn). ( d ) Principal component analysis (PCA) biplot of mean intensity of all tested surface markers with loading vectors (grey) of the five most prevalent markers. ( e ) PCA of marker intensity and morphological features color-coded based on Haasters’s classification . Green, blue, and red denote rapidly self-renewing (RS), spindle-shaped (SS), and flattened-cuboidal (FC) phenotyes, respectively. Relatively orthogonal loading vectors suggest low correlation between morphology and marker intensity. PCA was performed using OriginLab version 2019.

Article Snippet: Subfigures were generated using the OriginLab software version 2019 ( a – c ) and Fiji ImageJ software ( d ).

Techniques: Marker, Expressing, Single-cell Analysis, Standard Deviation

Journal: Scientific Reports

Article Title: Investigating heterogeneities of live mesenchymal stromal cells using AI-based label-free imaging

doi: 10.1038/s41598-021-85905-z

Figure Lengend Snippet: Spatial-temporal fluctuations of MSC marker expression. ( a ) Predicted time-lapse composite snapshots of a single cell at 0, 16, 32 and 48 h (left to right). ( b ) Total protein expression fluctuation for 4 surface markers (CD105, CD29, STRO-1, and CD44) of the predicted 48-h time-lapse video. Graph was created by using OriginLab version 2019. ( c ) Pairwise cross correlation between the tested markers. We calculated both the overall correlation (dark grey) and the fluctuation correlation (light grey) where the line thickness denotes the correlation magnitude. The fluctuation correlation (light grey) calculated using the single cell data is significantly lower than the overall correlation determined using all 500 cells, indicating remarkable stochasticity in the gene expression fluctuation. ( d ) Averaged autocorrelation functions (ACF) for all four tested markers exhibit a similar exponential decay. Inset shows the ACF decay rates, which are roughly the correlation time of them. ( d ) was created by using OriginLab version 2019.

Article Snippet: Subfigures were generated using the OriginLab software version 2019 ( a – c ) and Fiji ImageJ software ( d ).

Techniques: Marker, Expressing

Journal: Scientific Reports

Article Title: Investigating heterogeneities of live mesenchymal stromal cells using AI-based label-free imaging

doi: 10.1038/s41598-021-85905-z

Figure Lengend Snippet: Prediction of temporal marker expression holds information on intracellular heterogeneity. ( a , b ) Heatmaps showing intracellular marker distribution and fluctuation for typical MSC surface markers. Fluctuation patterns vary significantly between tested markers indicating strong marker dependency within a single cell. ( c ) Protein localization differences of CD105 and STRO-1 underlining marker dependency. ( d ) 2D Fast Fourier Transform of heatmaps ( a , b ) revealing oscillatory frequency differences in time and space. Red and black arrows highlighting characteristic temporal and spatial fluctuation lengths, respectively. CD105 shows highest temporal and spatial fluctuation. Subfigures were generated using the OriginLab software version 2019 ( a – c ) and Fiji ImageJ software ( d ).

Article Snippet: Subfigures were generated using the OriginLab software version 2019 ( a – c ) and Fiji ImageJ software ( d ).

Techniques: Marker, Expressing, Generated, Software