Identifying Medical Diagnoses and Treatable Diseases by Image-Based Deep Learning.

The implementation of clinical-decision support algorithms for medical imaging faces challenges with reliability and interpretability. Here, we establish a diagnostic tool based on a deep-learning framework for the screening of patients with common treatable blinding retinal diseases. Our framework utilizes transfer learning, which trains a neural network with a fraction of the data of conventional approaches. Applying this approach to a dataset of optical coherence tomography images, we demonstrate performance comparable to that of human experts in classifying age-related macular degeneration and diabetic macular edema. We also provide a more transparent and interpretable diagnosis by highlighting the regions recognized by the neural network. We further demonstrate the general applicability of our AI system for diagnosis of pediatric pneumonia using chest X-ray images. This tool may ultimately aid in expediting the diagnosis and referral of these treatable conditions, thereby facilitating earlier treatment, resulting in improved clinical outcomes. VIDEO ABSTRACT.

MISTERMINATE Mechanistically Links Mitochondrial Dysfunction with Proteostasis Failure.

Mitochondrial dysfunction and proteostasis failure frequently coexist as hallmarks of neurodegenerative disease. How these pathologies are related is not well understood. Here, we describe a phenomenon termed MISTERMINATE (mitochondrial-stress-induced translational termination impairment and protein carboxyl terminal extension), which mechanistically links mitochondrial dysfunction with proteostasis failure. We show that mitochondrial dysfunction impairs translational termination of nuclear-encoded mitochondrial mRNAs, including complex-I 30kD subunit (C-I30) mRNA, occurring on the mitochondrial surface in Drosophila and mammalian cells. Ribosomes stalled at the normal stop codon continue to add to the C terminus of C-I30 certain amino acids non-coded by mRNA template. C-terminally extended C-I30 is toxic when assembled into C-I and forms aggregates in the cytosol. Enhancing co-translational quality control prevents C-I30 C-terminal extension and rescues mitochondrial and neuromuscular degeneration in a Parkinson's disease model. These findings emphasize the importance of efficient translation termination and reveal unexpected link between mitochondrial health and proteome homeostasis mediated by MISTERMINATE.

Strain Solitons in an Epitaxially Strained van der Waals-like Material.

Strain solitons are quasi-dislocations that form in van der Waals materials to relieve the energy associated with lattice or rotational mismatch. Novel electronic properties of strain solitons were predicted and observed. To date, strain solitons have been observed only in exfoliated crystals or mechanically strained crystals. The lack of a scalable approach toward the generation of strain solitons poses a significant challenge in the study of and use of their properties. Here, we report the formation of strain solitons with epitaxial growth of bismuth on InSb(111)B by molecular beam epitaxy. The morphology of the strain solitons for films of varying thickness is characterized with scanning tunneling microscopy, and the local strain state is determined from atomic resolution images. Bending in the solitons is attributed to interactions with the interface, and large angle bending is associated with edge dislocations. Our results enable the scalable generation of strain solitons.

Quality-control mechanisms targeting translationally stalled and C-terminally extended poly(GR) associated with ALS/FTD.

Maintaining the fidelity of nascent peptide chain (NP) synthesis is essential for proteome integrity and cellular health. Ribosome-associated quality control (RQC) serves to resolve stalled translation, during which untemplated Ala/Thr residues are added C terminally to stalled peptide, as shown during C-terminal Ala and Thr addition (CAT-tailing) in yeast. The mechanism and biological effects of CAT-tailing-like activity in metazoans remain unclear. Here we show that CAT-tailing-like modification of poly(GR), a dipeptide repeat derived from amyotrophic lateral sclerosis with frontotemporal dementia (ALS/FTD)-associated GGGGCC (G4C2) repeat expansion in C9ORF72, contributes to disease. We find that poly(GR) can act as a mitochondria-targeting signal, causing some poly(GR) to be cotranslationally imported into mitochondria. However, poly(GR) translation on mitochondrial surface is frequently stalled, triggering RQC and CAT-tailing-like C-terminal extension (CTE). CTE promotes poly(GR) stabilization, aggregation, and toxicity. Our genetic studies in Drosophila uncovered an important role of the mitochondrial protease YME1L in clearing poly(GR), revealing mitochondria as major sites of poly(GR) metabolism. Moreover, the mitochondria-associated noncanonical Notch signaling pathway impinges on the RQC machinery to restrain poly(GR) accumulation, at least in part through the AKT/VCP axis. The conserved actions of YME1L and noncanonical Notch signaling in animal models and patient cells support their fundamental involvement in ALS/FTD.

Gate-tunable superconducting diode effect in a three-terminal Josephson device.

The phenomenon of non-reciprocal critical current in a Josephson device, termed the Josephson diode effect, has garnered much recent interest. Realization of the diode effect requires inversion symmetry breaking, typically obtained by spin-orbit interactions. Here we report observation of the Josephson diode effect in a three-terminal Josephson device based upon an InAs quantum well two-dimensional electron gas proximitized by an epitaxial aluminum superconducting layer. We demonstrate that the diode efficiency in our devices can be tuned by a small out-of-plane magnetic field or by electrostatic gating. We show that the Josephson diode effect in these devices is a consequence of the artificial realization of a current-phase relation that contains higher harmonics. We also show nonlinear DC intermodulation and simultaneous two-signal rectification, enabled by the multi-terminal nature of the devices. Furthermore, we show that the diode effect is an inherent property of multi-terminal Josephson devices, establishing an immediately scalable approach by which potential applications of the Josephson diode effect can be realized, agnostic to the underlying material platform. These Josephson devices may also serve as gate-tunable building blocks in designing topologically protected qubits.

Selective control of conductance modes in multi-terminal Josephson junctions.

The Andreev bound state spectra of multi-terminal Josephson junctions form an artificial band structure, which is predicted to host tunable topological phases under certain conditions. However, the number of conductance modes between the terminals of a multi-terminal Josephson junction must be few in order for this spectrum to be experimentally accessible. In this work, we employ a quantum point contact geometry in three-terminal Josephson devices to demonstrate independent control of conductance modes between each pair of terminals and access to the single-mode regime coexistent with the presence of superconducting coupling. These results establish a full platform on which to realize tunable Andreev bound state spectra in multi-terminal Josephson junctions.

Strain-Energy Release in Bent Semiconductor Nanowires Occurring by Polygonization or Nanocrack Formation.

Strain engineering of semiconductors is used to modulate carrier mobility, tune the energy bandgap, and drive growth of self-assembled nanostructures. Understanding strain-energy relaxation mechanisms including phase transformations, dislocation nucleation and migration, and fracturing is essential to both exploit this degree of freedom and avoid degradation of carrier lifetime and mobility, particularly in prestrained electronic devices and flexible electronics that undergo large changes in strain during operation. Raman spectroscopy, high-resolution transmission electron microscopy, and electron diffraction are utilized to identify strain-energy release mechanisms of bent diamond-cubic silicon (Si) and zinc-blende GaAs nanowires, which were elastically strained to >6% at room temperature and then annealed at an elevated temperature to activate relaxation mechanisms. High-temperature annealing of bent Si-nanowires leads to the nucleation, glide, and climb of dislocations, which align themselves to form grain boundaries, thereby reducing the strain energy. Herein, Si nanowires are reported to undergo polygonization, which is the formation of polygonal-shaped grains separated by grain boundaries consisting of aligned edge dislocations. Furthermore, strain is shown to drive dopant diffusion. In contrast to the behavior of Si, GaAs nanowires release strain energy by forming nanocracks in regions of tensile strain due to the weakening of As-bonds. These insights into the relaxation behavior of highly strained crystals can inform the design of nanoelectronic devices and provide guidance on mitigating degradation.

Charge Separation in Epitaxial SnS/MoS2 Vertical Heterojunctions Grown by Low-Temperature Pulsed MOCVD.

The weak van der Waals bonding between monolayers in layered materials enables fabrication of heterostructures without the constraints of conventional heteroepitaxy. Although many novel heterostructures have been created by mechanical exfoliation and stacking, the direct growth of 2D chalcogenide heterostructures creates new opportunities for large-scale integration. This paper describes the epitaxial growth of layered, p-type tin sulfide (SnS) on n-type molybdenum disulfide (MoS2) by pulsed metal-organic chemical vapor deposition at 180 °C. The influence of precursor pulse and purge times on film morphology establishes growth conditions that favor layer-by-layer growth of SnS, which is critical for materials with layer-dependent electronic properties. Kelvin probe force microscopy measurements determine a built-in potential as high as 0.95 eV, and under illumination a surface photovoltage is generated, consistent with the expected Type-II band alignment for a multilayer SnS/MoS2 heterostructure. The bottom-up growth of a nonisostructural heterojunction comprising 2D semiconductors expands the combinations of materials available for scalable production of ultrathin devices with field-tunable responses.

Grape skin extract improves muscle function and extends lifespan of a Drosophila model of Parkinson's disease through activation of mitophagy.

Recent studies suggest that moderate red wine consumption may confer several health benefits, including protection against heart disease, certain cancers and multiple age-related neurological diseases such as Alzheimer's disease. These health benefits are assumed to come from a compound from grape skin called resveratrol, which has been proposed to be a pro-longevity agent. Whether resveratrol accounts for all the health benefits of grape-derived nutrients and the molecular and cellular mechanisms underlying the beneficial effects of such nutrients are not well understood. Here we investigated the effect of supplementing grape skin extract (GSE) left from red wine-production process to the daily food intake of a Drosophila melanogaster model of Parkinson's disease (PD) associated with PTEN-induced kinase 1 (PINK1) loss-of-function. Consumption of GSE resulted in rescue of mitochondrial morphological defects, improvement of indirect flight muscle function and health-span, and prolonged lifespan of the PINK1 mutant flies. Further biochemical and genetic studies revealed a link between activation of mitophagy and the beneficial effects of GSE. Our results indicate that GSE can promote autophagy activation, preserve mitochondria function, and protect against PD pathogenesis, and that the beneficial effect of GSE in mitophagy activation is not accounted for by resveratrol alone.