Molecular states during acute COVID-19 reveal distinct etiologies of long-term sequelae

The post-acute sequelae of SARS-CoV-2 infection (PASC) affect a significant portion of COVID-19 survivors, leading to a wide range of persistent symptoms. While some studies have explored the immunological aspects of PASC, particularly its relationship to the acute phase of COVID-19, the sample sizes have often been small and molecular profiling limited. Previous research has yielded mixed results, with some showing no association between PASC and acute anti-spike antibody titers, while others identified a link between reduced total antibody titers and the development of PASC symptoms. Different subsets of PASC symptoms have been associated with various acute measures, including multi-omics data, viral presence, T cell phenotypes, and autoantibodies. Multiple hypotheses have been proposed to explain the connection between acute COVID-19 and PASC, such as chronic inflammation, autoimmunity, microbiome/virome dysbiosis, and persistent tissue damage. Despite these efforts, a comprehensive understanding of the heterogeneous molecular processes linking the acute host response to SARS-CoV-2 infection and the subsequent development of PASC remains elusive. This study aimed to address this gap by conducting a large-scale transcriptome-wide investigation of blood gene expression changes during acute COVID-19 and their association with subsequent PASC in a cohort of hospitalized patients with long-term follow-up.

Several studies have investigated the relationship between acute COVID-19 and the development of PASC. Some research has focused on the role of the immune response in predicting PASC outcomes. However, these studies often had small sample sizes and limited molecular profiling. Two recent studies utilized larger cohorts and broader molecular profiling, but their findings were somewhat contradictory. One study found no significant association between PASC and acute anti-spike antibody titers, while another revealed that decreased total antibody titers predicted PASC development. Other studies linked distinct PASC symptom subsets to acute measures from multi-omics data, presence of Epstein-Barr virus, specific T cell phenotypes, and autoantibodies. These studies highlighted the complexity of PASC and pointed towards various potential risk factors, but a more complete understanding of the molecular mechanisms was needed.

This study enrolled 567 individuals (495 hospitalized with COVID-19 and 72 controls) from the Mount Sinai COVID-19 Biobank Study. Blood samples were collected serially from hospitalized individuals and once from controls. RNA sequencing (RNA-seq) was performed on 1,392 samples. Six months or more after discharge, 232 individuals (165 with RNA-seq) completed a self-reported checklist assessing PASC symptoms. The researchers analyzed RNA-seq data from 361 acute blood samples (165 individuals) who had completed the PASC checklist. Cell type fractions were computationally estimated and validated using complete blood counts. Differential expression (DE) analyses were performed to identify acute gene expression patterns associating with PASC symptoms one year after discharge, controlling for several confounding variables. Cell type-specific (CTS) differences in gene expression were assessed using a DE model with an interaction term between symptom and estimated cell fraction. The association of CTS DEGs with anti-spike antibody titers was investigated by controlling for the titers in the DE analyses. The findings were validated in an independent dataset. ELISA was used to measure antibody titers, and Olink was used to analyze circulating proteins. The study also utilized extensive clinical data extracted from electronic health records and manual chart review.

The study found limited association between PASC symptoms and acute anti-spike antibody titers, which was validated in an independent dataset. However, distinct acute-phase cell-type-specific (CTS) gene expression signatures were identified that linked several immune cell types to post-acute sequelae. Plasma cells were implicated in the etiology of numerous symptoms. Two distinct clusters of PASC symptoms were identified based on shared plasma cell DE signatures. One cluster (plasma cell pulmonary cluster) was associated with downregulation of immunoglobulin-related genes, independent of anti-spike antibody titers. The other cluster (plasma cell miscellaneous cluster) showed upregulation of immunoglobulin-related genes, largely dependent on anti-spike antibody titers. The study also found associations between PASC symptoms and gene expression in other immune cell types (CD8+ and γδ T cells, memory resting CD4+ T cells, neutrophils, and memory-activated CD4+ T cells). Validation in an independent dataset confirmed the association between lower total immunoglobulins and PASC, independent of anti-spike antibody titers. The study's findings demonstrate the existence of at least two distinct etiologies for PASC, both detectable during acute COVID-19 and linked to the host response to the virus.

This study's large-scale transcriptome-wide analysis revealed that multiple etiologies of PASC are already detectable during acute COVID-19, directly linking PASC development to the acute host response. The identification of distinct gene expression signatures associated with different symptom clusters highlights the heterogeneity of PASC. The findings emphasize the importance of the acute phase of COVID-19 as a critical early window in the pathogenesis of PASC, warranting increased attention in future research designs. The contrasting gene expression patterns in the two plasma cell clusters, and their differing dependence on anti-spike antibody titers, provide strong evidence for at least two distinct etiologies of PASC. The observed downregulation of immunoglobulin-related genes in the pulmonary cluster, independent of anti-spike antibody titers, suggests a possible underlying non-specific downregulation of humoral immune activity. The validation of this finding in an independent dataset further strengthens this hypothesis. The association of gene expression in other immune cell types with PASC further underscores the complexity of PASC and the involvement of various immune mechanisms.

This study identified at least two distinct etiologies for different sets of PASC symptoms, one dependent and one independent of the antibody response to the SARS-CoV-2 spike protein. The association of acute-phase gene expression with PASC symptoms one year later establishes a direct link between the acute and post-acute phases. Future research should focus on capturing both acute and post-acute data in the same individuals to further elucidate the mechanisms driving PASC and to identify potential biomarkers for prediction and treatment. The study also highlights the need for more precise definitions of PASC phenotypes to enhance the power of future studies.

The study's reliance on self-reported PASC symptoms introduces potential bias. The exclusive inclusion of hospitalized patients limits the generalizability of the findings to milder cases or asymptomatic infections. The lack of post-acute molecular data prevents direct characterization of the connection between acute molecular signals and the molecular components of PASC. The use of a specific interaction model for identifying CTS DEGs requires further validation. The study was underpowered for some combinations of cell types and symptoms, indicating a need for larger cohorts in future research.