Differential roles of RIG-I like receptors in SARS-CoV-2 infection

Retinoic acid-inducible gene I (RIG-I) and melanoma differentiation-associated protein 5 (MDA5) sense viral RNA and activate antiviral immune responses. Herein we investigate their functions in human epithelial cells, the primary and initial target of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). A deficiency in MDA5, RIG-I or mitochondrial antiviral signaling protein (MAVS) enhanced viral replication. The expression of the type I/III interferon (IFN) during infection was impaired in MDA5−/− and MAVS−/−, but not in RIG-I−/−, when compared to wild type (WT) cells. The mRNA level of full-length angiotensin-converting enzyme 2 (ACE2), the cellular entry receptor for SARS-CoV-2, was ~ 2.5-fold higher in RIG-I−/− than WT cells. These data demonstrate MDA5 as the predominant SARS-CoV-2 sensor, IFN-independent induction of ACE2 and anti-SARS-CoV-2 role of RIG-I in epithelial cells. Supplementary Information The online version contains supplementary material available at 10.1186/s40779-021-00340-5.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is an enveloped, positive sense single-stranded RNA virus that has caused the greatest global public health crisis in the twenty-first century. Its pathogenesis remains largely unknown-highlighting a critical need for new research in this area. The cytoplasmic retinoic acid-inducible gene I (RIG-I) like receptors (RLRs) are major pattern recognition receptors (PRRs) for RNA viruses. Once engaged by viral RNA, RLRs bind mitochondrial antiviral signaling protein (MAVS), which ignites a signaling cascade, leading to transcription of immune genes [1]. Because of their importance to initiation of antiviral immune responses, these pathways are thus common targets of immune evasion by many viruses including SARS-CoV-2 [2].
We investigated the role of RLRs in controlling SARS-CoV-2 infection and mounting immune responses in a human lung epithelial cell line Calu-3. We generated individual knockouts using CRISPR-Cas9, and validated them by immunoblotting (Additional file 1: Fig. S1a). To prove that these gene functions are precisely silenced, we infected mutant cells with vesicular stomatitis virus (VSV, specifically activates RIG-I-MAVS) with a green fluorescence protein (GFP) integrated into its genome. As expected, RIG-I −/− or MAVS −/− cells presented a higher VSV-GFP load than wild type (WT) cells, while MDA5 −/− cells had a similar viral load as WT cells (Additional file 1: Fig. S1b). We then compared SARS-CoV-2 load and interferon (IFN) in these cells. The intracellular viral RNA loads were significantly higher in all knockout cells than WT cells at 24 and 72 h post infection (p.i.) (Additional file 2: Fig. S2a). Consistently, the extracellular viral titers produced by all knockout cells were also higher than those by WT cells (Additional file 2: Fig. S2b). We confirmed these observations in another human lung epithelial cell line A549 (Additional file 2: Fig. S2c), though which is significantly less permissive to SARS-CoV-2. Although primarily sensing DNA viruses, the cyclic GMP-AMP synthase (cGAS)-stimulator-of-interferon-genes (STING) signaling pathway also Open Access *Correspondence: pewang@uchc.edu Department of Immunology, School of Medicine, University of Connecticut Health Center, Farmington, CT 06030, USA restricts many RNA virus infection [3]. We noted a slight increase in SARS-CoV-2 load in STING −/− cells (Additional file 2: Fig. S2d, e), suggesting that STING signaling is largely dispensable for control of SARS-CoV-2.
We next examined antiviral immune responses. The IFNB1 (type I IFN) and IL29 (type III IFN) mRNA levels were continuously upregulated during the course of infection in WT cells; while this induction was impaired in MDA5 −/− and MAVS −/− cells, so was one of interferon-stimulated genes (ISG15) (Additional file 3: Fig. S3a). The concentrations of IFN-λ and C-X-C motif chemokine ligand 10 (CXCL10) proteins in the cell culture supernatants from MDA5 −/− and MAVS −/− were much lower than WT cells (Additional file 3: Fig. S3b). However, type I/III IFN and ISG15 expression was higher in RIG-I −/− than WT cells (Additional file 3: Fig. S3a), suggesting that RIG-I interferes with SARS-CoV-2 replication independently of IFNs. We next examined if RLR signaling regulates expression of angiotensin-converting enzyme 2 (ACE2), the predominant cellular entry receptor for SARS-CoV-2, thus influences viral replication. In the airway epithelium, in addition to full-length ACE2 (805 amino acids), a short isoform (459 amino acid) without the 17 aa of the signal peptide and 339 aa of the N-terminal peptidase domain is expressed. The short form, but not full-length, is inducible by type I/III IFNs. However, the short isoform fails to bind the SARS-CoV-2 spike protein, thus likely has no role in viral entry. We first quantitated full-length ACE2 using our own primers (targeting Exon 4 and 5) by quantitative RT-PCR. Of note, the mRNA level of full-length ACE2 was induced by over twofold in WT at 24 and 72 h when compared to 1 h p.i. It was also induced in MDA5 −/− and MAVS −/− as normally as in WT cells (Additional file 3: Fig. S3c), though these knockout cells were deficient in type I/III IFN expression (Additional file 3: Fig. S3a, b), suggesting that SARS-CoV-2 infection induces ACE2 expression in an IFN-independent manner. Intriguingly, it was ~ 2.5 fold higher in RIG-I −/− than WT cells throughout the course of infection (Additional file 3: Fig. S3c), suggesting that RIG-I might suppress full-length ACE2 transcription. We confirmed these results using a published primer pair for full-length ACE2 only (Additional file 3: Fig. S3d). We next assessed the expression of the short isoform with two unique pairs of primers according to recent studies, which designated it MIRb and dACE2 respectively. The short isoform was upregulated in WT cells by > 3.5 times at 72 h, when compared to 1 h p.i., however, it was not induced at all in RIG-I −/− cells (Additional file 3: Fig. S3d).
Understanding the major PRR pathways in the respiratory tract epithelial cells is physiologically meaningful as these cells are the first line of host defense. The RLR signaling is functional in all tissues and cell types, in contrast to viral RNA-sensing TLR3/7 that are primarily limited to immune cells. Our results demonstrate that MDA5 is the predominant RLR for SARS-CoV-2, consistent with two recent studies [4,5]. However, in neither MDA5 nor MAVS knockout cells, induction of IFNs was completely abolished, suggesting that other PRRs may collectively play a role. RIG-I deletion had no negative impact on IFN responses, but still enhanced viral replication, suggesting that RIG-I plays a MAVS-IFNindependent antiviral role. However, the role of RIG-I in SARS-CoV-2 infection is inconsistent. Yin et al. [4] demonstrated that RIG-I was dispensable for the control of SARS-CoV-2 replication, while both our and Yamada's [5] data suggested otherwise. Mechanistically, RIG-I likely binds the 3' untranslated region of the SARS-CoV-2 RNA genome via its helicase domains and prevents viral RNA replication independently of IFNs [5]. In addition to the above-mentioned mechanisms, our results suggest that RIG-I could restrain full-length ACE2 expression, consequently SARS-CoV-2 cellular entry. To our surprise, induction of the short isoform of ACE2 expression by SARS-CoV-2 seems dependent on RIG-I. Although the mechanism underlying the contrasting role of RIG-I in full-length/short ACE2 transcription remains unknown, notably, their transcription is indeed regulated differently. Comprehensive future work is necessary to elucidate this.
We want to point out that our findings are limited to human lung epithelial cell lines, and other PRRs such as viral RNA-sensing TLR3/7 may be important SARS-CoV-2 sensors in other cell types. Nonetheless, given the essential role of MDA5 in initiation of antiviral immune responses in the airway epithelium, the MDA5 agonists could thus be potentially therapeutic against early SARS-CoV-2 infection.