A potential role for AHR in SARS-CoV-2 pathology

Coronavirus infection is associated to life-threatening respiratory failure. The aryl hydrocarbon receptor (AHR) was recently identified as a host factor for Zika and dengue viruses; AHR antagonists decrease viral titers and ameliorate ZIKV-induced pathology in vivo. Here we report that AHR is activated during coronavirus infection, impacting anti-viral immunity and lung basal cells associated to tissue repair. Hence, AHR antagonists are candidate therapeutics for the management of coronavirus-infected patients.

response to M-CoV infection in vitro and in vivo 20 . We also detected the activation of AHR signaling in M-CoV 21 , HCoV-229E 22 , MERS-CoV 23 , SARS-CoV-1 20 and SARS-CoV-2 24 gene expression data available in the Gene Expression Omnibus (GEO) public database (Fig. 1a).
In depth analyses of RNA-Seq data from M-CoV infected bone marrow-derived macrophages detected the up-regulation of AHR and related genes such as IDO2, CYP1B1, AHRR and TIPARP (Fig 1b). IDO2 catalyzes the production of AHR agonists in the context of tumors 25 and viral infections 17,26 , and TIPARP contributes to the suppression of IFN-I expression 18 . Ingenuity pathway analysis (IPA) detected the enrichment of pattern recognition receptors and immune cell signaling molecules involved in antiviral IFN-related mechanisms, including NF-κB, JAK/Stat, PKR, IRF and IL-6, as well as a signi cant enrichment in AHR signaling (Fig. 1c). Moreover, upstream analysis identi ed AHR-ARNT as candidate regulators of the transcriptional response to M-CoV infection. These ndings suggest that AHR participates in the transcriptional response of M-CoV infected cells (Fig. 1d).
Next, we analyzed a dataset of HCoV-229E infected A549 cells; IPA analysis detected AHR among the most highly enriched pathways in infected cells. Moreover, we identi ed AHR as a regulator of the transcriptional response of infected samples (Figs. 1e,f). The analysis of RNA-seq data from MERS-CoV infected human lung adenocarcinoma (Calu-3) cells detected the up-regulation of AHR and related genes (CYPA1, CYP1B1 and TIPARP) (Fig. 1g). Accordingly, IPA detected the activation of a broad range of cellular processes, including AHR signaling (Fig. 1h). Of note, AHR, AHRR and CYP1A1 expression determined by RNA-seq was found to be gradually up-regulated at different times post infection (Fig. 1i).
Finally, we analyzed RNA-seq data of mock-infected and SARS-CoV-2 infected human primary lung epithelium cells. IPA of differentially expressed genes in SARS-CoV-2 infected cells compared to mockinfected cells detected the activation of the AHR pathway, together with IFN signaling, NF-KB, JAK/Stat and others (Figs. 1j,k). In addition, AHR was also identi ed as an upstream regulator of relevant cytokines and chemokines involved in the response to viral infection and in ammation-related cellular processes ( Fig. 1l).
We used a dataset of AHR targets identi ed in genome-wide ChIP-seq studies to de ne the AHRdependent module in the transcriptional response to coronavirus, focusing on M-CoV and MERS-CoV for which the available datasets were most complete (Figs. 2a,b). The pathway enrichment analysis of the AHR-dependent and the AHR-independent components of the transcriptional response to coronavirus infection detected an enrichment in biological pathways related to the immune response and Fc signaling in the AHR-dependent transcriptional module (Fig. 2c).
In about 5-15% of infected patients, SARS-CoV-2 infection causes life-threatening respiratory complications 10,11 . To investigate potential AHR-dependent mechanisms that may contribute to the pathogenesis of COVID-19, we analyzed a single-cell RNA-Seq (scRNA-seq) dataset of lung epithelial cells, identifying six cell populations corresponding to basal cells, goblet cells, ciliated cells, Tuft cells, neuroendocrine cells and pulmonary ionocytes (Fig. 2d). Strikingly, the AHR-dependent transcriptional module induced by coronavirus infection was mostly associated to basal cells (Fig. 2e), which contribute to lung regeneration after multiple types of injuries including in uenza infection 27 . Interestingly, IPA analysis of the scRNA-seq dataset of lung epithelial cells identi ed AHR as a transcriptional regulator of the basal cell cluster (Fig. 2f). These ndings suggest that AHR signaling triggered by coronavirus infection interferes with the regenerative activity of lung epithelial basal cells.
In summary, we identi ed AHR signaling as a common host response to infection by multiple coronaviruses. It has been reported that although some NF-κB signaling is needed for coronavirus replication, excessive activation of this pathway may be deleterious for the virus 22 . AHR limits NF-B activation, and interferes with multiple anti-viral immune mechanisms including IFN-I production and intrinsic immunity 17,18 . Thus, our ndings suggest that the modulation of NF-kB signaling via AHR may dampen the immune response against coronavirus. We also detected a potential role of AHR in the control Fc receptor expression and signaling. Based on recent reports on the association of high antibody titers against SARS-CoV-2 with worst clinical outcomes 28 , these ndings suggest a role for antibody enhancement in COVID-19 pathogenesis.
Our studies also suggest that AHR signaling associated to coronavirus infection affects lung basal cells, which give rise to stem cells involved in lung repair in multiple contexts including in uenza virus infection [29][30][31] . Of note, AHR-de cient mice show enhanced repair of the lung bronchiolar epithelium following naphthalene injury 32  Differential expression analysis. The count matrix was built using the Rsem output for each sample, and then DESeq2 38 was used to conduct differential expression analysis. The log2 fold change in the results was shrunk using Apeglm 39 .' Downstream analysis. Differentially expressed genes were further analyzed using GSEA 40 and IPA in order to nd enriched pathways and upstream regulators.
Overlap between the ChIP-seq and bulk RNA-sequencing results. The list of AHR target genes, generated from a ChIP-seq dataset 41 , was overlapped with the lists of differentially expressed genes from M-CoV and MERS-CoV-infected samples using the threshold of log2 fold change larger than 0 and adjusted p value smaller than 0.05. Then the results were further overlapped to generate a list of up-regulated genes common to all virus-infected samples and ChIP-seq identi ed AHR targets.
Download and processing of data. The single cell dataset was downloaded from the GEO repository (GSE103354). The 10X format count matrices were downloaded and processed using Seurat 42 including normalization, dimension reduction and clustering. Then for each cluster the GSEA was used to analyze the differential expression analysis results of each cluster with the AHR activation signature generated in the previously. The clusters that had signi cant up-regulation in AHR activation signature (FDR < 0.25) are marked as "AHR-dependent" and other clusters are marked as "AHR-independent". Then MAST 43 was used to conduct the differential expression analysis between the AHR activated population and non AHR activated population. Downstream analysis. Differentially expressed genes were further analyzed using GSEA and IPA in order to nd enriched pathways and upstream regulators.
Cell-trajectory and pseudo-time analysis. The single cell dataset was further analyzed using monocle 44 . The pseudo-time order of the cells were constructed using the genes that were differentially expressed (p adjusted value < 0.01) between the AHR and the non-AHR activated population.
Data availability. The authors declare that the raw data supporting the ndings of this study are publicly available. Transcriptomics data (RNA-Seq or microarray) from virus-infected samples, including M-CoV, HCoV-229E, MERS-CoV and SARS-CoV-2 were accessed at GSE144882, GSE89167, GSE139516, and GSE147507 respectively. Trachea epithelial single cell data was accessed at GSE103354.

Declarations
CONICET. We thank all members of the Garcia and Quintana laboratories for helpful advice and discussions.
Authors' contributions F.G and Z.L. performed bioinformatics analysis and/or interpreted ndings, F.G., Z.L., C.C.G. and F.J.Q. wrote the manuscript, C.C.G. and F.J.Q. designed and supervised the study and edited the manuscript.
Competing nancial interests FJQ is a member of the Scienti c Advisory Board of Kyn Therapeutics. The other authors declare no competing interests.

Figure 2
Identi cation of an AHR-dependent module in the transcriptional response to coronavirus infection. (a) Strategy used to identify AHR-dependent modules in the transcriptional response to coronavirus infection. Signi cantly up-regulated genes in each RNA-seq dataset were overlapped with genes associated to peaks identi ed by AHR ChIP-seq within 1 kilobase distance to the transcription start region. (b) Venn diagram representing the overlap in signi cantly up-regulated genes in M-CoV-infected cells, MERS-CoVinfected cells and the ones identi ed by an AHR ChIP-Seq experiment. (c) Pathway enrichment analysis was performed on two gene sets: the one resulting from the overlap of the three datasets in (b) ("AHR") and the one resulting from the overlap between the M-CoV and MERS-CoV datasets in (b) ("non-AHR"). p value was determined using a Fisher's exact test. Pathways colored in red are related to immunity (d) Single cell RNA-sequencing data of mouse trachea epithelia cells. Each cell is labelled by the cell type. (e) "AHR-dependent" and a "AHR-independent" cell populations identi ed by gene set enrichment analysis using the previously generated gene set (f) Upstream regulator analysis on the AHR activated cell population compared to the non-AHR activated cell population identi ed AHR as an upstream regulator.