ABSTRACT
The ongoing corona virus disease 2019 (COVID-19) pandemic has led to an urgent demand for appropriate models depicting host-pathogen interactions and disease severity-dependent immune responses. Amongst various animal models, hamster species are particularly valuable as they are permissive to develop a moderate (Mesocricetus auratus) or severe (Phodopus roborovskii) disease course following infection. Here, we use single-cell ribonucleic acid sequencing of white blood cells to dissect cell-specific changes in moderate and severe disease courses of hamsters infected with severe acute respiratory syndrome coronavirus 2. To determine universal and species-specific transcriptional responses, the generated datasets were integrated with two publicly available datasets of human COVID-19 patients (Schulte-Schrepping et al. 2020 and Su et al. 2020) featuring all disease severities. Datasets were integrated using the R package Harmony and the Python package scGen enabling the prediction of disease states through different species using an autoencoder neural network architecture. Specifically, application of a low dimensional latent space embedding allows capturing most relevant transcriptome data structures, identifying shift vectors from healthy to diseased cells as well as interspecies differences. Preliminary results show that interspecies integration of hamster and human data is achievable, and major cell types were identified throughout the datasets. Training of a neuronal network on human blood monocytes enables the prediction of transcriptomic disease severity specific patterns, paving the way for extended analyses involving several cell types and species. In addition to in-depth analysis of COVID-19 signatures in blood of hamsters and humans, successfully established workflows could subsequently be used to study the pathology of extensive lung diseases, shedding light on cellular mechanisms in the transition from healthy to diseased cellular states.