Effect of MR head coil geometry on deep-learning-based MR image reconstruction.

PURPOSE: To investigate whether parallel imaging-imposed geometric coil constraints can be relaxed when using a deep learning (DL)-based image reconstruction method as opposed to a traditional non-DL method. THEORY AND METHODS: Traditional and DL-based MR image reconstruction approaches operate in fundamentally different ways: Traditional methods solve a system of equations derived from the image data whereas DL methods use data/target pairs to learn a generalizable reconstruction model. Two sets of head coil profiles were evaluated: (1) 8-channel and (2) 32-channel geometries. A DL model was compared to conjugate gradient SENSE (CG-SENSE) and L1-wavelet compressed sensing (CS) through quantitative metrics and visual assessment as coil overlap was increased. RESULTS: Results were generally consistent between experiments. As coil overlap increased, there was a significant (p < 0.001) decrease in performance in most cases for all methods. The decrease was most pronounced for CG-SENSE, and the DL models significantly outperformed (p < 0.001) their non-DL counterparts in all scenarios. CS showed improved robustness to coil overlap and signal-to-noise ratio (SNR) versus CG-SENSE, but had quantitatively and visually poorer reconstructions characterized by blurriness as compared to DL. DL showed virtually no change in performance across SNR and very small changes across coil overlap. CONCLUSION: The DL image reconstruction method produced images that were robust to coil overlap and of higher quality than CG-SENSE and CS. This suggests that geometric coil design constraints can be relaxed when using DL reconstruction methods.

Short-term repeatability and long-term reproducibility of quantitative MR imaging biomarkers in a single centre longitudinal study.

Quantitative imaging biomarkers (QIBs) can be defined as objective measures that are sensitive and specific to changes in tissue physiology. Provided the acquired QIBs are not affected by scanner changes, they could play an important role in disease diagnosis, prognosis, management, and treatment monitoring. The precision of selected QIBs was assessed from data collected on a 3-T scanner in four healthy participants over a 5-year period. Inevitable scanner changes and acquisition protocol revisions occurred during this time. Standard and custom processing pipelines were used to calculate regional brain volume, cortical thickness, T2, T2*, quantitative susceptibility, cerebral blood flow, axial, radial and mean diffusivity, peak width of skeletonized mean diffusivity, and fractional anisotropy from the acquired images. Coefficient of variation (CoV) and intra-class correlation (ICC) indices were determined in the short-term (i.e., repeatable over three acquisitions within 4 weeks) and in the long-term (i.e., reproducible over four acquisition sessions in 5 years). Precision indices varied based on acquisition technique, processing pipeline, and anatomical region. Good repeatability (average CoV=2.40% and ICC=0.78) and reproducibility (average CoV=8.86 % and ICC=0.72) were found over all QIBs. The best performance indices were obtained for diffusion derived biomarkers (CoV∼0.96% and ICCs=0.87); conversely, the poorest indices were found for the cerebral blood flow biomarker (CoV>10% and ICC<0.5). These results demonstrate that changes in protocol, along with hardware and software upgrades, did not affect the estimates of the selected biomarkers and their precision. Further characterization of the QIB is necessary to understand meaningful changes in the biomarkers in longitudinal studies of normal brain aging and translation to clinical research.

Interaction of endogenous tau protein with synaptic proteins is regulated by N-methyl-D-aspartate receptor-dependent tau phosphorylation.

Amyloid-ß and tau protein are the two most prominent factors in the pathology of Alzheimer disease. Recent studies indicate that phosphorylated tau might affect synaptic function. We now show that endogenous tau is found at postsynaptic sites where it interacts with the PSD95-NMDA receptor complex. NMDA receptor activation leads to a selective phosphorylation of specific sites in tau, regulating the interaction of tau with Fyn and the PSD95-NMDA receptor complex. Based on our results, we propose that the physiologically occurring phosphorylation of tau could serve as a regulatory mechanism to prevent NMDA receptor overexcitation.

Hyperphosphorylation and cleavage at D421 enhance tau secretion.

It is well established that tau pathology propagates in a predictable manner in Alzheimer's disease (AD). Moreover, tau accumulates in the cerebrospinal fluid (CSF) of AD's patients. The mechanisms underlying the propagation of tau pathology and its accumulation in the CSF remain to be elucidated. Recent studies have reported that human tau was secreted by neurons and non-neuronal cells when it was overexpressed indicating that tau secretion could contribute to the spreading of tau pathology in the brain and could lead to its accumulation in the CSF. In the present study, we showed that the overexpression of human tau resulted in its secretion by Hela cells. The main form of tau secreted by these cells was cleaved at the C-terminal. Surprisingly, secreted tau was dephosphorylated at several sites in comparison to intracellular tau which presented a strong immunoreactivity to all phospho-dependent antibodies tested. Our data also revealed that phosphorylation and cleavage of tau favored its secretion by Hela cells. Indeed, the mimicking of phosphorylation at 12 sites known to be phosphorylated in AD enhanced tau secretion. A mutant form of tau truncated at D421, the preferential cleavage site of caspase-3, was also significantly more secreted than wild-type tau. Taken together, our results indicate that hyperphosphorylation and cleavage of tau by favoring its secretion could contribute to the propagation of tau pathology in the brain and its accumulation in the CSF.

Increased association between rough endoplasmic reticulum membranes and mitochondria in transgenic mice that express P301L tau.

In several neurodegenerative diseases, including Alzheimer disease, the neuronal microtubule-associated protein tau becomes hyperphosphorylated, accumulates in the somatodendritic compartment, and aggregates into insoluble filaments. The consequences of the accumulation of hyperphosphorylated tau in the somatodendritic compartment remain poorly characterized at the early stage of disease before the formation of tau insoluble filaments. We investigated the ultrastructural changes induced by this accumulation in the neuronal soma of motor neurons in asymptomatic JNPL3 mice that overexpress mutant tau, P301L. More numerous contacts between rough endoplasmic reticulum (RER) membranes and mitochondria were observed in JNLP3 mice compared with wild-type mice. This correlated with a preferential increase of the amount of tau at the surface of RER membranes but not at the surface of mitochondria, as revealed by tau immunogold labeling. Using a subcellular fractionation procedure, an increased amount of phosphorylated tau was identified in the rough microsome subfraction, wherein the RER marker, ribophorin, was enriched. A similar increase was noted in the rough microsome subfraction isolated from Alzheimer disease brains. The association of hyperphosphorylated tau with ER membranes was confirmed by double immunogold labeling of the subfraction enriched in ER membranes isolated from Alzheimer disease brains. These results suggest that more contacts between RER membranes and mitochondria resulting from the accumulation of tau at the surface of RER membranes might contribute to tau-induced neurodegeneration.

Tau interacts with Golgi membranes and mediates their association with microtubules.

Tau, a microtubule-associated protein enriched in the axon, is known to stabilize and promote the formation of microtubules during axonal outgrowth. Several studies have reported that tau was associated with membranes. In the present study, we further characterized the interaction of tau with membranous elements by examining its distribution in subfractions enriched in either Golgi or endoplasmic reticulum membranes isolated from rat brain. A subfraction enriched with markers of the medial Golgi compartment, MG160 and mannosidase II, presented a high tau content indicating that tau was associated with these membranes. Electron microscope morphometry confirmed the enrichment of this subfraction with Golgi membranes. Double-immunogold labeling experiments conducted on this subfraction showed the direct association of tau with vesicles labeled with either an antibody directed against MG160 or TGN38. The association of tau with the Golgi membranes was further confirmed by immunoisolating Golgi membranes with an anti-tau antibody. Immunogold labeling confirmed the presence of tau on the Golgi membranes in neurons in vivo. Overexpression of human tau in primary hippocampal neurons induced the formation of large Golgi vesicles that were found in close vicinity to tau-containing microtubules. This suggested that tau could serve as a link between Golgi membranes and microtubules. Such role for tau was demonstrated in an in vitro reconstitution assay. Finally, our results showed that some tau isoforms present in the Golgi subfraction were phosphorylated at the sites recognized by the phosphorylation-dependent antibodies PHF-1 and AT-8.

Interaction of microtubule-associated protein-2 and p63: a new link between microtubules and rough endoplasmic reticulum membranes in neurons.

Neurons are polarized cells presenting two distinct compartments, dendrites and an axon. Dendrites can be distinguished from the axon by the presence of rough endoplasmic reticulum (RER). The mechanism by which the structure and distribution of the RER is maintained in these cells is poorly understood. In the present study, we investigated the role of the dendritic microtubule-associated protein-2 (MAP2) in the RER membrane positioning by comparing their distribution in brain subcellular fractions and in primary hippocampal cells and by examining the MAP2-microtubule interaction with RER membranes in vitro. Subcellular fractionation of rat brain revealed a high MAP2 content in a subfraction enriched with the endoplasmic reticulum markers ribophorin and p63. Electron microscope morphometry confirmed the enrichment of this subfraction with RER membranes. In cultured hippocampal neurons, MAP2 and p63 were found to concomitantly compartmentalize to the dendritic processes during neuronal differentiation. Protein blot overlays using purified MAP2c protein revealed its interaction with p63, and immunoprecipitation experiments performed in HeLa cells showed that this interaction involves the projection domain of MAP2. In an in vitro reconstitution assay, MAP2-containing microtubules were observed to bind to RER membranes in contrast to microtubules containing tau, the axonal MAP. This binding of MAP2c microtubules was reduced when an anti-p63 antibody was added to the assay. The present results suggest that MAP2 is involved in the association of RER membranes with microtubules and thereby could participate in the differential distribution of RER membranes within a neuron.

The projection domain of MAP2b regulates microtubule protrusion and process formation in Sf9 cells.

The expression of microtubule-associated protein 2 (MAP2), developmentally regulated by alternative splicing, coincides with neurite outgrowth. MAP2 proteins contain a microtubule-binding domain (C-terminal) that promotes microtubule assembly and a poorly characterized domain, the projection domain (N-terminal), extending at the surface of microtubules. MAP2b differs from MAP2c by an additional sequence of 1372 amino acids in the projection domain. In this study, we examined the role of the projection domain in the protrusion of microtubules from the cell surface and the subsequent process formation in Sf9 cells. In this system, MAP2b has a lower capacity to induce process formation than MAP2c. To investigate the role of the projection domain in this event, we expressed truncated forms of MAP2b and MAP2c that have partial or complete deletion of their projection domain in Sf9 cells. Our results indicate that process formation is induced by the microtubule-binding domain of these MAP2 proteins and is regulated by their projection domain. Furthermore, the microtubule-binding activity of MAP2b and MAP2c truncated forms as well as the structural properties of the microtubule bundles induced by them do not seem to be the only determinants that control the protrusion of microtubules from the cell surface in Sf9 cells. Rather, our data suggest that microtubule protrusion and process formation are regulated by intramolecular interactions between the projection domain and its microtubule-binding domain in MAP2b.