Benchmarking of deep learning algorithms for 3D instance segmentation of confocal image datasets.

Segmenting three-dimensional (3D) microscopy images is essential for understanding phenomena like morphogenesis, cell division, cellular growth, and genetic expression patterns. Recently, deep learning (DL) pipelines have been developed, which claim to provide high accuracy segmentation of cellular images and are increasingly considered as the state of the art for image segmentation problems. However, it remains difficult to define their relative performances as the concurrent diversity and lack of uniform evaluation strategies makes it difficult to know how their results compare. In this paper, we first made an inventory of the available DL methods for 3D cell segmentation. We next implemented and quantitatively compared a number of representative DL pipelines, alongside a highly efficient non-DL method named MARS. The DL methods were trained on a common dataset of 3D cellular confocal microscopy images. Their segmentation accuracies were also tested in the presence of different image artifacts. A specific method for segmentation quality evaluation was adopted, which isolates segmentation errors due to under- or oversegmentation. This is complemented with a 3D visualization strategy for interactive exploration of segmentation quality. Our analysis shows that the DL pipelines have different levels of accuracy. Two of them, which are end-to-end 3D and were originally designed for cell boundary detection, show high performance and offer clear advantages in terms of adaptability to new data.

Fast-field-cycling NMR at very low magnetic fields: water molecular dynamic biomarkers of glioma cell invasion and migration.

Our objective was to study NMR relaxometry of glioma invasion/migration at very low field (<2 mT) by fast-field-cycling NMR (FFC-NMR) and to decipher the pathophysiological processes of glioma that are responsible for relaxation changes in order to open a new diagnostic method that can be extended to imaging. The phenotypes of two new glioma mouse models, Glio6 and Glio96, were characterized by T2w -MRI, HE histology, Ki-67 immunohistochemistry (IHC) and CXCR4 RT-qPCR, and were compared with the U87 model. R1 dispersions of glioma tissues were acquired at low field (0.1 mT-0.8 T) ex vivo and were fitted with Lorentzian and power-law models to extract FFC biomarkers related to the molecular dynamics of water. In order to decipher relaxation changes, three main invasion/migration pathophysiological processes were studied: hypoxia, H2 O2 function and the water-channel aquaporin-4 (AQP4). Glio6 and Glio96 were characterized with invasion/migration phenotype and U87 with high cell proliferation as a solid glioma. At very low field, invasion/migration versus proliferation was characterized by a decrease in the relaxation-rate constant (ΔR1 ≈ -32% at 0.1 mT) and correlation time (≈-40%). These decreases corroborated the AQP4-IHC overexpression (Glio6/Glio96: +92%/+46%), suggesting rapid transcytolemmal water exchange, which was confirmed by the intracellular water-lifetime τIN decrease (ΔτIN ≈ -30%). In functional experiments, AQP4 expression, τIN and the relaxation-rate constant at very low field were all found to be sensitive to hypoxia and to H2 O2 stimuli. At very low field the role of water exchanges in relaxation modulation was confirmed, and for the first time it was linked to the glioma invasion/migration and to its main pathophysiological processes: hypoxia, H2 O2 redox signaling and AQP4 expression. The method appears appropriate to evaluate the effect of drugs that can target these pathophysiological mechanisms. Finally, FFC-NMR operating at low field is demonstrated to be sensitive to invasion glioma phenotype and can be straightforwardly extended to FFC-MRI as a new cancer invasion imaging method in the clinic.

CEST-MRI for glioma pH quantification in mouse model: Validation by immunohistochemistry.

In glioma, the acidification of the extracellular tumor microenvironment drives proliferation, angiogenesis, immunosuppression, invasion and chemoresistance. Therefore, quantification of glioma extracellular pH (pHe) is of crucial importance. This study is focused on the application of the YbHPDO3A (ytterbium 1,4,7-triscarboxymethyl-1,4,7,10-tetraazacyclododecane) probe for in vivo glioma pHe quantification using chemical exchange saturation transfer (CEST)-MRI and its correlation with tumor metabolism assessed by immunohistochemistry. The U87 glioma mouse model was used (n = 18) and MRI performed at 4.7 T. CEST-MRI of YbHPDO3A solutions at different pH values showed two resolved CEST spectra at 71 ppm and 99 ppm, both sensitive to pH variations, allowing therefore calculation of the ratiometric curve for in vivo pH quantification. In vivo MRI sequences consisted of T2w for tumor localization, T2w * to assess YbHPDO3A biodistribution by exploiting its magnetic susceptibility effect and CEST for glioma pHe mapping. T2w * images show that YbHPDO3A extravasates in tumor in regions with damaged blood-brain barrier. The pHe is calculated only in these regions. Hematoxylin/eosin histology and Ki-67, CA-IX (carbonic anhydrase 9) and NHE-1 immunohistochemical staining were performed; their expression rates were compared with the in vivo pHe values. On the basis of the cell proliferation marker Ki-67, two groups were defined: one group with a lower mitotic index (MI% < 20% = mean value) and a mean pHe value of 7.00 (low-proliferation/high-pH group) and the other with MI% > 20% and an acidic pHe of 6.6 (high-proliferation/low-pH group). CA-IX and NHE-1 were over-expressed in the high-proliferation/low-pH group (CA-IX, 92 ± 7% versus 30 ± 13%; NHE-1, 84 ± 8% versus 35 ± 11%), indicating an acidic/hypoxic microenvironment. These immunohistochemical results are consistent with our pHe mapping (Pearson correlation coefficient > 0.70) and provide evidence for the feasibility of the CEST-MRI method with the YbHPDO3A probe for glioma pHe quantification at 4.7 T. Importantly, the YbHPDO3A probe has similar chemical and biological properties to the clinically approved MRI contrast agent GdHPDO3A. This makes the method promising for a clinical translation.

Growth directions and stiffness across cell layers determine whether tissues stay smooth or buckle.

From smooth to buckled, nature exhibits organs of various shapes and forms. How cellular growth patterns produce smooth organ shapes such as leaves and sepals remains unclear. Here we show that unidirectional growth and comparable stiffness across both epidermal layers of Arabidopsis sepals are essential for smoothness. We identified a mutant with ectopic ASYMMETRIC LEAVES 2 (AS2) expression on the outer epidermis. Our analysis reveals that ectopic AS2 expression causes outer epidermal buckling at early stages of sepal development, due to conflicting growth directions and unequal epidermal stiffnesses. Aligning growth direction and increasing stiffness of the outer epidermis restores smoothness. Furthermore, buckling influences auxin efflux transporter protein PIN-FORMED 1 polarity to generate outgrowth in the later stages, suggesting that buckling is sufficient to initiate outgrowths. Our findings suggest that in addition to molecular cues influencing tissue mechanics, tissue mechanics can also modulate molecular signals, giving rise to well-defined shapes.