Mutations accumulated in the Spike of SARS-CoV-2 Omicron allow for more efficient counteraction of the restriction factor BST2/Tetherin

BST2/Tetherin is a restriction factor with broad antiviral activity against enveloped viruses, including coronaviruses. Specifically, BST2 traps nascent particles to membrane compartments, preventing their release and spread. In turn, viruses have evolved multiple mechanisms to counteract BST2. Here, we examined the interactions between BST2 and SARS-CoV-2. Our study shows that BST2 reduces SARS-CoV-2 virion release. However, the virus uses the Spike (S) protein to downregulate BST2. This requires a physical interaction between S and BST2, which routes BST2 for lysosomal degradation in a Clathtin- and ubiquitination-dependent manner. By surveying different SARS-CoV-2 variants of concern (Alpha-Omicron), we found that Omicron is more efficient at counteracting BST2, and that mutations in S account for its enhanced anti-BST2 activity. Mapping analyses revealed that several surfaces in the extracellular region of BST2 are required for an interaction with the Spike, and that the Omicron variant has changed its patterns of association with BST2 to improve its counteraction. Therefore, our study suggests that, besides enhancing receptor binding and evasion of neutralizing antibodies, mutations accumulated in the Spike afford more efficient counteraction of BST2, which highlights that BST2 antagonism is important for SARS-CoV-2 infectivity and spread.

On the Importance of Attention and Augmentations for Hypothesis Transfer in Domain Adaptation and Generalization

Unsupervised domain adaptation (UDA) aims to mitigate the performance drop due to the distribution shift between the training and testing datasets. UDA methods have achieved performance gains for models trained on a source domain with labeled data to a target domain with only unlabeled data. The standard feature extraction method in domain adaptation has been convolutional neural networks (CNNs). Recently, attention-based transformer models have emerged as effective alternatives for computer vision tasks. In this paper, we benchmark three attention-based architectures, specifically vision transformer (ViT), shifted window transformer (SWIN), and dual attention vision transformer (DAViT), against convolutional architectures ResNet, HRNet and attention-based ConvNext, to assess the performance of different backbones for domain generalization and adaptation. We incorporate these backbone architectures as feature extractors in the source hypothesis transfer (SHOT) framework for UDA. SHOT leverages the knowledge learned in the source domain to align the image features of unlabeled target data in the absence of source domain data, using self-supervised deep feature clustering and self-training. We analyze the generalization and adaptation performance of these models on standard UDA datasets and aerial UDA datasets. In addition, we modernize the training procedure commonly seen in UDA tasks by adding image augmentation techniques to help models generate richer features. Our results show that ConvNext and SWIN offer the best performance, indicating that the attention mechanism is very beneficial for domain generalization and adaptation with both transformer and convolutional architectures. Our ablation study shows that our modernized training recipe, within the SHOT framework, significantly boosts performance on aerial datasets.

Arrayed Imaging Reflectometry monitoring of anti-viral antibody production throughout vaccination and breakthrough Covid-19

Immune responses to COVID-19 infection and vaccination are individual and varied. There is a need to understand the timeline of vaccination efficacy against current and yet to be discovered viral mutations. Assessing immunity to SARS-CoV-2 in the context of immunity to other respiratory viruses is also valuable. Here we demonstrate the capability of a fully automated prototype Arrayed Imaging Reflectometry system to perform reliable longitudinal serology against a 34-plex respiratory array. The array contains antigens for respiratory syncytial virus, seasonal influenza, common human coronaviruses, MERS, SARS-CoV-1, and SARS-CoV-2. AIR measures a change in reflectivity due to the binding of serum antibodies to the antigens on the array. Samples were collected from convalescent COVID-19 donors and individuals vaccinated with a two-dose mRNA vaccine regimen. Vaccinated samples were collected prior to the first dose, one week after the first dose, one week after the second dose, and monthly thereafter. Information following booster dose and/or breakthrough infection is included for a subset of subjects. Longitudinal samples of vaccinated individuals demonstrate a rise and fall of SARS-CoV-2 spike antibodies in agreement with general knowledge of the adaptive immune response and other studies. Linear Regression analysis was performed to understand the relationship between antibodies binding to different antigens on the array. Our analysis identified strong correlations between closely related influenza virus strains as well as correlations between SARS-CoV-2, SARS-CoV-1, and human coronavirus 229E. A small test of using diluted whole blood from a fingerstick provided clean arrays with antibody binding comparable to serum. Potential applications include assessing immunity in the context of exposure to multiple respiratory viruses, clinical serology, population monitoring to facilitate public health recommendations, and vaccine development against new viruses and virus mutations.