ABSTRACT:

Infection with SARS-CoV-2, the virus that causes COVID-19, can lead to severe lower respiratory illness including pneumonia and acute respiratory distress syndrome, which can result in profound morbidity and mortality. However, many infected individuals are either asymptomatic or have isolated upper respiratory symptoms, which suggests that the upper airways represent the initial site of viral infection, and that some individuals are able to largely constrain viral pathology to the nasal and oropharyngeal tissues. Which cell types in the human nasopharynx are the primary targets of SARS-CoV-2 infection, and how infection influences the cellular organization of the respiratory epithelium remains incompletely understood. Here, we present nasopharyngeal samples from a cohort of 35 individuals with COVID-19, representing a wide spectrum of disease states from ambulatory to critically ill, as well as 23 healthy and intubated patients without COVID-19. Using standard nasopharyngeal swabs, we collected viable cells and performed single-cell RNA-sequencing (scRNA-seq), simultaneously profiling both host and viral RNA. We find that following infection with SARS-CoV-2, the upper respiratory epithelium undergoes massive reorganization: secretory cells diversify and expand, and mature epithelial cells are preferentially lost. Further, we observe evidence for deuterosomal cell and immature ciliated cell expansion, potentially representing active repopulation of lost ciliated cells through coupled secretory cell differentiation. Epithelial cells from participants with mild/moderate COVID-19 show extensive induction of genes associated with anti-viral and type I interferon responses. In contrast, cells from participants with severe lower respiratory symptoms appear globally muted in their anti-viral capacity, despite substantially higher local inflammatory myeloid populations and equivalent nasal viral loads. This suggests an essential role for intrinsic, local epithelial immunity in curbing and constraining viral-induced pathology. Using a custom computational pipeline, we characterized cell-associated SARS-CoV-2 RNA and identified rare cells with RNA intermediates strongly suggestive of active replication. Both within and across individuals, we find remarkable diversity and heterogeneity among SARS-CoV-2 RNA+ host cells, including developing/immature and interferon-responsive ciliated cells, KRT13+ “hillock”-like cells, and unique subsets of secretory, goblet, and squamous cells. Finally, SARS-CoV-2 RNA+ cells, as compared to uninfected bystanders, are enriched for genes involved in susceptibility (e.g., CTSL, TMPRSS2) or response (e.g., MX1, IFITM3, EIF2AK2) to infection. Together, this work defines both protective and detrimental host responses to SARS-CoV-2, determines the direct viral targets of infection, and suggests that failed anti-viral epithelial immunity in the nasal mucosa may underlie the progression to severe COVID-19.

BRIEF SUMMARY OF COHORT AND SINGLE CELL TRANSCRIPTOMICS METHODS:

Nasopharyngeal swabs were collected from 58 individuals from the University of Mississippi Medical Center between April and September 2020. A Control cohort consisted of 15 individuals who were asymptomatic and SARS-CoV-2 negative by PCR, 2 individuals were asymptomatic, SARS-CoV-2 negative by PCR, but had recent COVID-19 (asymptomatic for > 40 days), and 6 intubated individuals in the intensive care unit who were SARS-CoV-2 negative and with no recent history of COVID-19. 35 individuals were diagnosed with COVID-19, and nasopharyngeal swabs were collected within the first 3 days following admission to the hospital. Using the World Health Organization (WHO) guidelines for stratification and classification of COVID-19 severity based on the level of required respiratory support, 14 of the individuals were considered COVID-19 mild/moderate (WHO score 1-5) and 21 had severe COVID-19 (WHO score 6-8). Samples from the nasopharyngeal epithelium were taken by a trained healthcare provider and rapidly cryopreserved to maintain cellular viability. Swabs were later processed to recover single-cell suspensions (mean +/- SEM: 57,000 +/- 15,000 total cells recovered per swab), before generating single-cell transcriptomes via Seq-Well S^3 platform. Libraries were generated using Illumina Nextera XT Library Prep Kits and sequenced on NextSeq 500/550 High Output v2.5 kits to an average depth of 180 million aligned reads per array: read 1: 21 (cell barcode, UMI), read 2: 50 (digital gene expression), index 1: 8 (N700 barcode).  Pooled libraries were demultiplexed using bcl2fastq (v2.17.1.14) with default settings (mask_short_adapter_reads 10, minimum_trimmed_read_length 10, implemented using Cumulus, snapshot 4, https://cumulus.readthedocs.io/en/stable/bcl2fastq.html). Libraries were aligned using STAR within the Drop-Seq Computational Protocol (https://github.com/broadinstitute/Drop-seq) and implemented on Cumulus (https://cumulus.readthedocs.io/en/latest/drop_seq.html, snapshot 9, default parameters). A custom reference was created by combining human GRCh38 (from CellRanger version 3.0.0, Ensembl 93) and SARS-CoV-2 RNA genomes. The SARS-CoV-2 viral sequence and GTF are as described in Kim et al. 2020 (https://github.com/hyeshik/sars-cov-2-transcriptome, BetaCov/South Korea/KCDC03/2020 based on NC_045512.2). The GTF includes all CDS regions (as of this annotation of the transcriptome, the CDS regions completely cover the RNA genome without overlapping segments), and regions were added to describe the 5’ UTR (“SARSCoV2_5prime”), the 3’ UTR (“SARSCoV2_3prime”), and reads aligning to anywhere within the Negative Strand (“SARSCoV2_NegStrand”). Trailing A’s at the 3’ end of the virus were excluded from the SARS-CoV-2 FASTA, as these were found to drive spurious viral alignment in pre-COVID19 samples. Finally, additional small sequences were appended to the FASTA and GTF that differentiate reads that align to the 70-nucleotide region around the viral TRS sequence - either across the intact, unspliced genomic sequences (e.g. named “SARSCoV2_Unspliced_S” or “SARSCoV2_Unspliced_Leader”) or various spliced RNA species (e.g. “SARSCoV2_Spliced_Leader_TRS_S”). Alignment references were tested against a diverse set of pre-COVID-19 samples and in vitro SARS-CoV-2 infected human bronchial epithelial cultures (Ravindra et al. bioRxiv 2020) to confirm specificity of viral aligning reads. For analysis of RNA velocity, we also recovered both exonic and intronic alignment information using DropEst (Cumulus (https://cumulus.readthedocs.io/en/latest/drop_seq.html, snapshot 9, dropest_velocyto true, run_dropest true). Aligned cell-by-gene matrices were merged across all study participants, and cells were filtered to eliminate barcodes with fewer than 200 UMI, 150 unique genes, and greater than 50% mitochondrial reads (cutoffs determined by distributions of reads across cells,). This resulted in a final dataset of 32,871 genes and 32,588 cells across 58 study participants, with an average recovery of 562 +/- 69 cells per swab (mean +/- SEM). 

PREPRINT: Ziegler et al. bioRxiv, 2021 https://www.biorxiv.org/content/10.1101/2021.02.20.431155v1

Correspondence to: Jose Ordovas-Montanes (jose.ordovas-montanes@childrens.harvard.edu), Bruce Horwitz (bruce.horwitz@childrens.harvard.edu), Sarah C. Glover (scglover@umc.edu), and Alex K. Shalek (shalek@mit.edu)