Integrin α₅β₁ contributes to cell fusion and inflammation mediated by SARS-CoV-2 spike via RGD-independent interaction

Significance

SARS-CoV-2 uses human ACE2 as the cell surface receptor to infect host cells, ultimately leading to the development of COVID-19. The complex progression of the disease and its impact on multiple organs suggest that additional cellular factors may be involved in the interaction with SARS-CoV-2. Integrin α₅β₁, a cell adhesion molecule that is widely expressed in various tissues, has emerged as one such factor. Our study has revealed that SARS-CoV-2 spike protein exploits α₅β₁ signaling to facilitate cell–cell fusion and trigger inflammatory responses through the interaction with α₅β_1. Both processes may contribute to the infection and pathogenesis of SARS-CoV-2.

Abstract

The Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus infects host cells by engaging its spike (S) protein with human ACE2 receptor. Recent studies suggest the involvement of integrins in SARS-CoV-2 infection through interaction with the S protein, but the underlying mechanism is not well understood. This study investigated the role of integrin α₅β₁, which recognizes the Arg-Gly-Asp (RGD) motif in its physiological ligands, in S-mediated virus entry and cell–cell fusion. Our results showed that α₅β₁ does not directly contribute to S-mediated cell entry, but it enhances S-mediated cell–cell fusion in collaboration with ACE2. This effect cannot be inhibited by the putative α₅β₁ inhibitor ATN-161 or the high-affinity RGD-mimetic inhibitor MK-0429 but requires the participation of α₅ cytoplasmic tail (CT). We detected a direct interaction between α₅β₁ and the S protein, but this interaction does not rely on the RGD-containing receptor binding domain of the S1 subunit of the S protein. Instead, it involves the S2 subunit of the S protein and α₅β₁ homo-oligomerization. Furthermore, we found that the S protein induces inflammatory responses in human endothelial cells, characterized by NF-κB activation, gasdermin D cleavage, and increased secretion of proinflammatory cytokines IL-6 and IL-1β. These effects can be attenuated by the loss of α₅ expression or inhibition of the α₅ CT binding protein phosphodiesterase-4D (PDE4D), suggesting the involvement of α₅ CT and PDE4D pathway. These findings provide molecular insights into the pathogenesis of SARS-CoV-2 mediated by a nonclassical RGD-independent ligand-binding and signaling function of integrin α₅β₁ and suggest potential targets for antiviral treatment.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the virus that causes COVID-19, uses the same cell entry receptor, angiotensin-converting enzyme 2 (ACE2), as SARS-CoV. However, it has a broader cell tropism, higher transmission rates, and a more complex pathogenesis, suggesting that additional host factors and mechanisms specific to SARS-CoV-2 are involved (1, 2). The severity of COVID-19 is associated with dysregulated inflammatory responses, characterized by elevated levels of cytokines such as IL-6 and IL-1β, as well as injury to vascular endothelial cells (3, 4). Beyond respiratory epithelial cells primarily targeted by SARS-CoV-2, a wide range of cell types in vascular, immune, and even central nervous systems experience dysfunction, contributing to the pathogenesis of COVID-19 (5–7). While innate immune receptors such as toll-like receptors (TLRs) may play a role in COVID-19 inflammation (8), they function as intrinsic receptors against most bacterial and viral infections. The specific host receptors responsible for the intricated pathogenesis of COVID-19 have yet to be fully delineated.

SARS-CoV-2 uses its spike (S) protein to bind to ACE2 for cell entry (9). Recent studies have indicated potential roles of integrins in SARS-CoV-2 infection (10–16) and COVID-19 pathogenesis (17–23). In humans, there are 24 α/β integrin heterodimers formed by the combination of 18 α and 8 β subunits (24). Integrins play critical roles in various biological processes, including leukocyte recirculation and migration, wound healing, blood clotting, and immune response (24). Some viruses use integrins as receptors or coreceptors to facilitate infection (25–27). The consideration of integrins as putative receptors for SARS-CoV-2 was primarily based on the presence of an integrin-binding RGD (Arg-Gly-Asp) motif on the SARS-CoV-2 S (SARS2-S) protein (28, 29). Among the eight integrins that recognize the RGD motif, α₅β₁, α_Vβ₃, and α_IIbβ₃ have been reported to directly interact with the SARS2-S protein in studies using purified proteins (12, 14, 17–19). However, the results of these studies have shown inconsistencies and controversies. Some indirect evidence for the interaction between integrins and SARS2-S was obtained from cell-based assays such as cell adhesion (19, 30). Of note, the potential ligand-binding inhibitor ATN-161 for α₅β₁ has been used in SARS-CoV-2 infection and cellular function assays (10, 20, 31). However, ATN-161 is not a direct RGD-blocking inhibitor, and its mechanism of inhibition remains undefined.

The objective of this study was to investigate the potential role of α₅β₁ in SARS-CoV-2 spike-mediated infection, its interaction with SARS2-S, and the resulting cellular response. Our data showed that α₅β₁ does not play a direct role in virus cell entry, but instead facilitates cell–cell fusion mediated by the SARS2-S protein. We also detected a direct interaction between α₅β₁ and SARS2-S, which does not depend on the RGD motif. This interaction involves the non-receptor-binding S2 subunit of SARS2-S and the homo-oligomerization of α₅β₁. Furthermore, our study demonstrated that SARS2-S induces proinflammatory responses in human endothelial cells, evidenced by NF-κB activation, cytokine release, and cleavage of gasdermin D (GSDMD). These processes were found to be mediated by α₅β₁ signaling through the PDE4D pathway. α₅β₁ is widely expressed in immune cells, lung, heart, and endothelial cells. The SARS2-S-induced and α₅β₁-mediated cellular responses may contribute to the complex pathogenesis of COVID-19.

Results

Integrin α₅β₁ Does Not Play a Role in S-Mediated Virus Entry.

To investigate the involvement of α₅β₁ in SARS2-S mediated cell entry, we used S-pseudotyped lentivirus and a replication-competent recombinant vesicular stomatitis virus (rVSV) encoding enhanced green fluorescent protein (EGFP) and SARS2-S protein (rVSV-S) or VSV G protein (rVSV-G) for virus infection assays. HEK293T-ACE2 stable cells naturally express high levels of α₅β₁, which was knocked out (KO) using CRISPR/cas9 (Fig. 1A). Lentivirus and rVSV-S infection were assessed by measuring luciferase activity and EGFP expression, respectively. Notably, the absence of α₅ expression in HEK293T-ACE2 cells had no impact on SARS2-S-pseudotyped lentivirus infection, similar to the control lentivirus pseudotyped with VSV G or SARS-CoV S (SARS-S) protein (Fig. 1B). Additionally, overexpression of α₅β₁ in HEK293T-ACE2-α₅-KO cells also did not affect lentivirus infection (Fig. 1B). When ACE2 was absent, rVSV-S failed to infect HEK293T cells (Fig. 1 C and D), despite their high expression of α₅β₁.

Fig. 1.

Integrin α₅β₁ does not contribute to the cell entry mediated by SARS-CoV-2 spike. (A) Surface expression of α₅ and ACE2 in HEK293T-ACE2 stable cells with or without α₅ knockout. (B) The HEK293T-ACE2, HEK293T-ACE2-α₅-KO, or HEK293T-ACE2-α₅-KO overexpressing α₅ were infected for 48 h with VSV-G, SARS-S, or SARS2-S-pseudotyped lentiviruses. The cells were harvested for luciferase activity measurement using Bright-Glo Luciferase Assay System. Data are mean ± SD from four independent repeats. (C) HEK293T cells transfected with or without ACE2 were infected by recombinant VSV virus carrying SARS2-S and EGFP for 24 h. Representative live-cell images from three independent repeats with three images randomly taken for each well were acquired using AMG EVOS fluorescence microscope with Plan Fluor 2× objective lens (numerical aperture of 0.06), equipped with Sony 1cx285AQ color charge-coupled device camera (CCD). (D) Quantification of rVSV-S virus infection based on cell images in panel (C). Data are mean ± SD from seven randomly taken images of three independent repeats. (E and F) Flow cytometry and immunoblot analysis of Vero E6 cells with and without α₅ knockout. (G) Representative images of Vero WT or Vero α₅-KO cells infected by rVSV-S or rVSV-G virus for 6 h. The images were acquired using EVOS M7000 imaging system with Plan Fluor 4× objective lens (numerical aperture of 0.13). Ten to Twenty images were randomly taken for each repeat in three independent repeats. (H) Quantification of virus infection results of panel (G). The infection was measured as number of EGFP-positive objects, mean EGFP area, or total EGFP area. Data are mean ± SD. Unpaired two-tailed t test. (I) HEK293T-ACE2 or HEK293T-ACE2-α₅-KO cells were infected for 6 h with rVSV-S in the absence or presence of 0.1 mM ATN-161 or pooled sera of COVID-19 vaccinated donors. The infection was measured by flow cytometry. Data are mean ± SD from five independent repeats.

To validate these findings, we performed additional experiments using Vero E6 cells that express high levels of endogenous ACE2. The α₅ in Vero E6 cells was knocked out using CRISPR/cas9, as demonstrated by flow cytometry and western blot analysis (Fig. 1 E and F). Following infection with rVSV-S or the control rVSV-G virus for 6 h, no discernible differences in the number of infected cells were observed between wild-type Vero E6 and Vero E6 α₅-KO cells (Fig. 1 G and H). These results indicate that α₅β₁ does not contribute to the S-mediated cell entry of SARS-CoV-2.

We examined the effect of ATN-161 on rVSV-S infection of HEK293T-ACE2 cells. Interestingly, the inhibitor exhibited no inhibition effect on rVSV-S infection in either wild-type or α₅-KO HEK293T-ACE2 cells, whereas pooled sera from COVID-19 vaccinated donors completely blocked virus infection (Fig. 1I).

Integrin α₅β₁ Contributes to S-Mediated Cell–Cell Fusion.

The expression of the S protein on cell surface induces cell–cell fusion (32), which facilitates the spread of virus among cells (33). To assess the extent of cell–cell fusion following rVSV-S infection in Vero E6 cells, we measured the mean or total EGFP areas. The results showed that loss of α₅ expression in Vero E6 cells significantly reduced cell–cell fusion induced by the rVSV-S virus, while it had no effect on cell–cell fusion induced by the rVSV-G virus (Fig. 1 G and H).

To further investigate a potential role of α₅β₁ in S-mediated cell–cell fusion, we used a well-established split GFP assay (32). The large GFP1-10 and small GFP11 fragments were expressed separately in HEK293T-α₅-KO cells along with ACE2 and SARS2-S, respectively. The fusion of ACE2 and S cells leads to the reconstitution of GFP fluorescence (Fig. 2A). Coexpression of S and GFP11, with or without α₅β₁, showed comparable levels of S surface expression (Fig. 2B). In the absence of ACE2, no cell–cell fusion was observed (Fig. 2C). When ACE2 and S cells were cocultured for 6 h, substantial cell–cell fusion occurred, as indicated by GFP fluorescence (Fig. 2D). Quantitative analysis of cell–cell fusion, based on mean or total GFP area, showed that coexpression of S and α₅β₁ significantly increased cell–cell fusion at both 6 h and 24 h (Fig. 2E). To examine whether the signaling function of α₅ is involved in this process, we truncated the cytoplasmic tail (CT) of α₅ at the conserved GFFKR motif. Remarkably, coexpression of α₅-CTtr/β₁ with the S protein failed to enhance cell–cell fusion (Fig. 2E), despite α₅-CTtr/β₁ and α₅/β₁ exhibiting comparable expression levels (Fig. 2B).

Fig. 2.

Integrin α₅β₁ contributes to cell–cell fusion mediated by SARS-CoV-2 spike. (A) Diagram of cell–cell fusion assay measured by split GFP assay. (B) Surface expression of SARS2-S and α₅ in HEK293T-α₅-KO cells. Cells were transfected with indicated constructs for 48 h and subjected to flow cytometry before the cell–cell fusion assay. Data are mean ± SD from four independent repeats. (C) α₅β₁ alone in the absence of ACE2 does not mediate cell–cell fusion induced by SARS2-S. Representative images of three independent repeats. (D) Representative images of HEK293T-α₅-KO cells transfected with ACE2 plus GFP1-10 cocultured for 6 h with HEK293T-α₅-KO cells transfected with S/GFP11, α₅β₁/S/GFP11, or α₅CTtrβ₁/S/GFP11. The α₅β₁/S/GFP11 cells were also pretreated with ATN-161 or MK-0429 before coculturing with the ACE2/GFP1-10 cells. Cells were imaged using AMG EVOS fluorescence microscope with Plan Fluor 4× objective lens (numerical aperture of 0.13), equipped with Sony 1cx285AQ color CCD camera. (E and F) Quantification of cell–cell fusion at 6 h and 24 h after coculturing. Three images were randomly taken for each experiment group in three independent repeats. The mean or total EGFP area was measured using CellProfiler software. Data are mean ± SD. One-Way ANOVA Tukey’s multiple comparisons test. *P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001.

Additionally, we found that ATN-161 and the high-affinity RGD-mimetic inhibitor MK-0429 had no effect on S-mediated cell–cell fusion in the presence of α₅β₁ (Fig. 2F). These results demonstrate the role of α₅β₁ in promoting S-mediated cell–cell fusion, which does not depend on the RGD-binding function but rather requires the participation of α₅ CT.

Integrin α₅β₁ Interacts with the S Protein of SARS-CoV-2 Independently of the RGD Motif.

The S protein is composed of S1 and S2 subunits (Fig. 3A), with the S1 subunit containing the receptor binding domain (RBD) responsible for ACE2 binding, and the S2 subunit containing the machinery for membrane fusion (Fig. 3A). The structures of S homotrimer (Fig. 3A) and the complex of RBD bound with ACE2 have been determined (Fig. 3B) (34, 35). The putative integrin binding RGD motif is located in an α-helix of the RBD at the ACE2 binding interface (Fig. 3B). To investigate the interaction between α₅β₁ and S protein, we used purified ectodomains of SARS2-S and SARS-S, the RBD of SARS2-S, and Fc-tagged S1 and S2 subunits of SARS2-S (Fig. 3C). All proteins were verified to be free of aggregates using size-exclusion chromatography (SEC). enzyme-linked immunosorbent assay (ELISA) was employed to detect the direct interaction of SARS2-S and α₅β₁ on the plate coated with equal molar concentrations of SARS2-S and RBD. A concentration-dependent binding of α₅β₁ to SARS2-S was detected by anti-α₅ mAb VC5 (Fig. 3D). However, while purified ACE2 bound well to RBD, no binding was detected between α₅β₁ and RBD coated at 2 or 15 μg/mL (Fig. 3D). Consistently, only soluble SARS2-S but not the RBD inhibited the binding of α₅β₁ with immobilized SARS2-S (Fig. 3E). Additionally, our ELISA result showed that α₅β₁ bound to SARS2-S significantly better than SARS-S (Fig. 3F), whereas ACE2 bound equally well to both proteins (Fig. 3F). Furthermore, our pull-down assay using protein A beads showed that α₅β₁ associated with SARS2-S even in the presence of an abundance of ACE2-Fc (Fig. 3G), suggesting that α₅β₁ and ACE2 can bind simultaneously to SARS2-S. This result contradicts the RGD-dependent binding model of α₅β₁ with SARS2-S since the location of the RGD motif on the RBD implies that the binding of α₅β₁ and ACE2 to the RBD is mutually exclusive (Fig. 3B).

Fig. 3.

Integrin α₅β₁ interacts with SARS2-S protein independent of the RGD motif. (A) The domain organization and cryo-EM structure of SARS2-S (PDB 6XR8). NTD: N-terminal domain; RBD: receptor binding domain; FP: fusion peptide; HR: heptad repeat; CD: C-terminal domain; TM: transmembrane. The putative N-glycan sites are marked by triangles. The surface exposed residues that are different between SARS-S and SARS2-S S2 subunits are shown as red sticks. (B) Crystal structure of SARS2-S RBD in complex with ACE2 (PDB 6LZG). The RGD motif is shown as green sticks. Disulfide bonds are red sticks. (C) Reducing SDS-PAGE of purified spike proteins. The recombinant stabilized SARS2-S or SARS-S ectodomain was expressed as secreted form in HEK293 or CHO cells. The SARS2-S RBD domain was expressed in HEK293T cells as a secreted form. The SARS2-S S1 and S2 subunits were expressed as secreted form with a C-terminal human IgG Fc tag in HEK293T cells. (D) Interaction of α₅β₁ with SARS2-S detected by ELISA. The plate was coated with equal molar concentration of S (15 μg/mL) and RBD (2 μg/mL), or RBD at a higher concentration (15 μg/mL). The binding of α₅β₁ was detected by mAb VC5. Binding of ACE2-Fc at 0.5 and 2.5 μg/ml to RBD was as a control. Data are mean ± SD from three or four independent repeats. (E) Competition ELISA. α₅β₁ was preincubated with either SARS2-S or RBD at different molar ratio before adding to SARS2-S-coated ELISA plate. Data are mean ± SD from three independent repeats. (F) ELISA plate was coated with 15 μg/mL of SARS-S, SARS2-S, or bovine serum albumin (BSA). Binding of α₅β₁ at 25 μg/mL or ACE2-Fc at 0.75 μg/mL was measured. Data are mean ± SD from four independent repeats. One-Way ANOVA Tukey’s multiple comparisons test. (G) α₅β₁ coimmunoprecipitation with SARS2-S and ACE2. Purified α₅β₁ ectodomain (6×His tag on β₁) was mixed with purified ACE2-Fc in the presence or absence of purified 6×His-tagged stabilized SARS2-S ectodomain and pulled down with protein A beads. The samples were first immunoblotted with anti-His and then reblotted with anti-Fc. (H) Pull-down assay of full-length α₅β₁ and S proteins. HEK293T-α₅-KO cells were transfected with α₅-EGFP/β₁ or EGFP plus C-terminal C9-tagged full-length SARS-S, SARS2-S, or SARS2-S with RGD to RGA mutation. The cell lysates were immunoprecipitated with GFP-Trap beads. The precipitated samples were immunoblotted with anti-C9 and reblotted with anti-EGFP under reducing condition. The intact α₅-EGFP and α₅ light chain tagged with EGFP were detected. (I) The RGD-mimetic α₅β₁ inhibitor MK-0429 (IC₅₀ 12 nM) had no effect on α₅β₁ binding with SARS2-S. HEK293T-α₅-KO cells were transfected with α₅-EGFP/β₁ or EGFP plus C-terminal PC-tagged full-length SARS2-S. The cells were treated with MK-0429 at 10 μM for 1 h before pull-down with GFP-Trap beads. (J and K) The purified Flag-tagged α₅β₁ (J) or α₈β₁ (K) was incubated with SARS2-S S1-Fc or S2-Fc and precipitated by protein A beads. The precipitated samples were immunoblotted with anti-Flag and anti-Fc under reducing condition.

The pentapeptide ATN-161 (Ac-PHSCN-NH₂) has been widely used as an α₅β₁ antagonist. It was designed based on the synergy sequence PHSRN of fibronectin domain 9 (Fn9) with an R to C substitution (SI Appendix, Fig. S1A) (36). A structural study of α₅β₁ bound with the Fn7-10 fragment revealed the high-resolution interaction between the synergy site and α₅ subunit (37), emphasizing the critical role of the R residue within the PHSRN sequence in interacting with α₅ (SI Appendix, Fig. S1A). The R to C substitution in ATN-161 is likely to reduce its interaction with α₅ if it binds to the synergy binding site. Our ELISA results showed no significant inhibitory effect of ATN-161 on the interaction between α₅β₁ and SARS2-S (SI Appendix, Fig. S1B).

Using a pull-down assay with GFP-Trap beads, we observed that α₅-EGFP/β₁ precipitated the full-length SARS2-S but not SARS-S in HEK293T-α₅-KO cells (Fig. 3H and SI Appendix, Fig. S2A). Mutating the RGD motif to RGA had no effect on the association of full-length SARS2-S with α₅-EGFP/β₁ (Fig. 3H). Furthermore, the RGD-mimetic inhibitor MK-0429 did not impede the pull-down of SARS2-S with α₅-EGFP/β₁ (Fig. 3I). We also introduced Ala mutations for two conserved Ser residues (S132 and S134) in the β₁ subunit, which disrupt RGD binding to α₅β₁ by interfering with metal ion binding (SI Appendix, Fig. S2B). Our pull-down assay showed that the β₁-SSAA mutation had no impact on the association of α₅-EGFP/β₁ with full-length SARS2-S (SI Appendix, Fig. S2C). Additionally, truncation of the α₅ CT did not affect the pull-down of full-length SARS2-S with α₅-CTtr/β₁-GFP (SI Appendix, Fig. S2D). We also found that α₅β₁ could associate with purified or surface-expressed S proteins of Delta and Omicron variants of SARS-CoV-2 (SI Appendix, Fig. S2 E and F). These results collectively demonstrate an RGD-independent interaction between the S protein of SARS-CoV-2 and α₅β₁.

We further investigated the interaction between α₅β₁ and SARS2-S using purified Fc-tagged S1 and S2 subunits of SARS2-S along with the Flag-tagged α₅β₁ ectodomain (SI Appendix, Fig. S2G). The pull-down results using protein A beads revealed the association of α₅β₁ with S2-Fc but not S1-Fc (Fig. 3J and SI Appendix, Fig. S2H). As a control, neither S1-Fc nor S2-Fc captured the Flag-tagged α₈β₁ (Fig. 3K). As expected, only S1-Fc but not S2-Fc bound the Flag-tagged ACE2 (SI Appendix, Fig. S2I). Additionally, no interaction between α₅β₁ and the S2-Fc of SARS-S was detected (SI Appendix, Fig. S2J). These data provide evidence for a direct interaction between SARS2-S and integrin α₅β₁, which is independent of the RGD-containing RBD but relies on the S2 subunit of SARS2-S.

The SARS2-S and α₅β₁ Interaction Involves α₅β₁ Homo-Oligomerization.

During the purification of the α₅β₁ ectodomain using high-performance microscale SEC, we observed five elution peaks (Fig. 4A), likely corresponding to five distinct populations migrating differently under native condition on blue native-PAGE (Fig. 4A). Although the 12 factions collected based on five SEC peaks were indistinguishable on reducing SDS-PAGE under denatured condition (Fig. 4B), they exhibited different migration patterns on blue native-PAGE (Fig. 4C). These protein bands migrating differently on blue native-PAGE were confirmed to be α₅β₁ proteins as determined by anti-α₅ and anti-β₁ western blotting (Fig. 4D).

Fig. 4.

The SARS2-S and α₅β₁ interaction involves α₅β₁ homo-oligomerization. (A) Blue native-PAGE and microscale size exclusion chromatography (SEC) of affinity-purified α₅β₁ ectodomain. Twelve fractions were collected for the 5 SEC peaks. (B–D) Reducing SDS-PAGE, blue native-PAGE, and immunoblot of the 12 α₅β₁ SEC fractions in A. (E) Hydrodynamic radii of the α₅β₁ fractions measured by dynamic light scattering (DLS) and their binding with stabilized SARS2-S ectodomain measured by ELISA. (F) DLS analysis of α₅β₁ SEC fractions corresponding to peak 1 and peaks 4 to 5 in the absence or presence of SARS2-S protein. Data are mean ± SD from five independent measurements. (G) Model of FKBP12-mediated homodimerization of α₅β₁. FKBP12 was tagged to the cytoplasmic tail (CT) of α₅. Binding of the bivalent FKBP12 ligand B/B homodimerizer induces dimerization of α₅β₁. (H and I) B/B increased the association of α₅-FKBP12/β₁-GFP determined by pull-down assay. HEK293T-α₅-KO cells were transfected with α₅-FKBP12/β₁-GFP plus C9-tagged full-length SARS2-S for 48 h. The cells were treated with or without B/B at different concentrations for 1 h before pull-down with GFP-Trap beads. The pull-down samples were subjected to immunoblot with anti-C9 and anti-α₅ antibodies. The immunoblot was quantified by S signal as a percentage of α₅ signal. Data are mean ± SD from two independent repeats. (J) Cell–cell fusion assay. HEK293T-α₅-KO cells transfected with α₅-FKBP12/β₁ and full-length SARS2-S plus GFP11 were treated with or without B/B and cocultured for 6 h with HEK293T-α₅-KO cells transfected ACE2/GFP1-10. The cell fusion was measured as total EGFP area and normalized to the area without B/B. Data are mean ± SD from three independent repeats with three randomly taken images for each repeat. Unpaired two-tailed t test. *P < 0.05.

We then used dynamic light scattering (DLS) to measure the hydrodynamic radii of the SEC fractions, revealing a positive correlation between particle sizes (in diameter) and the SEC retention volumes (in fraction numbers) (Fig. 4E). Intriguingly, our ELISA results revealed that SARS2-S specifically bound to α₅β₁ fractions with larger diameters but not those with smaller diameter (Fig. 4E). This observation was further validated by DLS assay, which indicated the formation of a complex between SARS2-S with α₅β₁ fractions from peak-1 but not peaks 4 to 5 (Fig. 4F).

Given that the molecular weight (~500 kDa) and diameter (~25 nm) of α₅β₁ peak-1 fractions closely resemble those of an α₅β₁ homodimer, we hypothesized that α₅β₁ homodimerization might be involved in its interaction with SARS2-S. To test this hypothesis, we introduced a FKBP12 tag to the CT of full-length α₅, which undergoes homodimerization upon binding the bivalent ligand B/B homodimerizer (Fig. 4G). Our pull-down results showed that the presence of B/B markedly enhanced the association between full-length SARS2-S and α₅-FKBP12/β₁-GFP in HEK293T-α₅-KO cells (Fig. 4 H and I). Furthermore, B/B also significantly increased cell–cell fusion between HEK293T-α₅-KO cells expressing α₅-FKBP12/β₁ plus SARS2-S and the cells expressing ACE2 (Fig. 4J). These results strongly suggest the involvement of α₅β₁ homodimerization in the interaction with SARS2-S.

The S Protein of SARS-CoV-2 Induces Inflammatory Responses Mediated by α₅β₁ and cAMP Regulation in Endothelial Cells.

To investigate the potential cellular response resulting from the interaction between SARS2-S and α₅β₁, we used an immortalized human umbilical vein endothelial cell line (HUVEC) as a model system. We employed CRISPR/Cas9 to knock out the endogenous α₅ expression in HUVEC, as demonstrated by flow cytometry and immunoblot analysis (Fig. 5A). Given that α₅β₁ has been implicated in endothelial inflammation (22, 38), we evaluated NF-κB activation and IL-6 secretion in HUVEC cells attached to plastic surface coated with purified S proteins. Compared to the poly-lysine control, we observed a significant increase in NF-κB phosphorylation in cells adhered to SARS2-S (Fig. 5 B and C and SI Appendix, Fig. S2 A–C). Remarkably, cells adhered to SARS-S or the RBD and S1 subunit of SARS2-S showed little to no NF-κB phosphorylation (Fig. 5 B and C and SI Appendix, Fig. S2 A–C). However, robust NF-κB phosphorylation was observed in cells adhered to the S2 subunit of SARS2-S (Fig. 5 B and C and SI Appendix, Fig. S2 A–C). The loss of α₅ expression significantly reduced SARS2-S-induced NF-κB activation (Fig. 5 B and C and SI Appendix, Fig. S2 A–C). Consistent with the NF-κB activation, both SARS2-S and its S2 subunit significantly increased IL-6 secretion in HUVEC cells, whereas SARS-S and the RBD and S1 subunit of SARS2-S had no such effect (Fig. 5D). α₅ knock-out significantly diminished SARS2-S-induced IL-6 release, but it had much less effect on LPS-induced IL-6 release (Fig. 5E). These findings demonstrate that SARS2-S, specifically through its S2 subunit, induces an inflammatory response in endothelial cells that is dependent on α₅β₁.

Fig. 5.

SARS2-S induces inflammatory signaling through α₅β₁ integrin in HUVECs. (A) Loss of α₅ expression in HUVEC-α₅-KO cells. The depletion of α₅ expression in HUVEC cells shown by flow cytometry (Left) and immunoblot (Right). (B) NF-κB activation in HUVEC cells attached to immobilized S proteins. HUVEC or HUVEC-α₅-KO cells were seeded into a plate coated with 10 μg/mL indicated proteins for 2 h. Total cells were collected for Jess automatic western blot. (C) Quantification of immunoblot data of panel (B). One represented data of four repeats is shown. (D) SARS2-S-induced IL-6 release of HUVEC cells. HUVEC cells were seeded into a plate coated with 10 μg/mL indicated proteins for 2 h. The IL-6 concentrations in the supernatants were measured by ELISA. Data are mean ± SD from four independent repeats. One-Way ANOVA Tukey’s multiple comparisons test. *P < 0.05; ns P > 0.05. (E) Loss of α₅ significantly reduced SARS2-S- but not LPS-induced IL-6 release in HUVEC cells. Two-tailed unpaired Welch’s t test. Data are mean ± SD from four independent repeats. (F) Dose-dependent inhibition by rolipram of SARS2-S-induced IL-6 release in HUVEC cells. HUVEC cells were treated with different concentrations of rolipram before seeding into a plate coated with SARS2-S or LPS. Data are mean ± SD from three or four independent repeats. (G) Dose-dependent inhibition by Dibutyryl-cAMP (D-cAMP) of SARS2-S-induced IL-6 release in HUVEC cells. Data are mean ± SD from three or four independent repeats. (H and I) SARS2-S- or LPS-induced cleavage of GSDMD in HUVEC detected by immunoblot in total cell lysates (H). The cleavage of GSDMD was presented as a percent of cleaved GSDMD to total GSDMD (I). (J) Loss of α₅ greatly reduced SARS2-S- but not LPS-induced GSDMD cleavage in HUVEC cells. For (I) and (J), Data are mean ± SD from two independent repeats.

Previous studies have demonstrated that α₅ integrin, through its cytoplasmic interaction with phosphodiesterase-4 (PDE4) isoform D5, plays a role in promoting inflammatory signaling in endothelial cells (38). To investigate the involvement of PDE4 in the SARS2-S-induced inflammatory response, we used a selective PDE4 inhibitor called rolipram (39). A dose–response assay showed that rolipram efficiently blocked SARS2-S-induced IL-6 release, while its impact on LPS-induced IL-6 release was comparably less pronounced (Fig. 5F). Active PDE4 is known to modulate inflammation by degrading cAMP (40), and a decrease in cAMP levels triggers the activation of inflammasome (41). Consistent with this scenario, the SARS2-S-induced IL-6 release was inhibited by elevating intracellular cAMP levels using cell membrane permeable Dibutyryl-cAMP (D-cAMP) (Fig. 5G). Additionally, SARS2-S stimulation led to the cleavage of gasdermin D (GSDMD) (Fig. 5 H and I), a hallmark of inflammasome activation. Knockout of α₅ integrin markedly attenuated SARS2-S-indcued GSDMD cleavage, while its impact on LPS-induced GSDMD cleavage was minimal (Fig. 5J). These results indicate that the inflammatory response induced by the interaction between SARS2-S and α₅β₁ involves the α₅-PDE4D-cAMP pathway.

To further validate these findings, we used HULEC-5a, a human-lung-derived microvascular endothelial cell line, which provides a more relevant model for studying lung microvascular injury associated with SARS-CoV-2 infection (42). We exposed HULEC-5a cells to soluble SARS2-S, RBD, or LPS in the absence or presence of D-cAMP. Consistent with our observations in HUVECs, treatment with SARS2-S, but not its RBD, led to NF-κB activation in HULEC-5a cells, and this response was attenuated by D-cAMP (Fig. 6 A and B). Moreover, SARS2-S, but not its RBD, triggered the cleavage of GSDMD (Fig. 6 C and D), which was also inhibited by D-cAMP (Fig. 6 C and D). Additionally, SARS2-S, but not its RBD, induced the secretion of IL-6 and IL-1β, which were effectively blocked by D-cAMP (Fig. 6E). Both IL-1β release and GSDMD cleavage are well-recognized indicators of inflammasome activation (43). As a control, D-cAMP reduced the NF-κB activation, GSDMD cleavage, IL-6 and IL-1β release induced by LPS (Fig. 6 A–E), consistent with the well-established anti-inflammatory function of cAMP.

Fig. 6.

SARS2-S induces inflammatory response in HULECs. (A and B) Soluble SARS2-S-induced NF-κB activation and inhibition by D-cAMP. HULEC-5a cells in suspension were incubated with 10 μg/mL SARS2-S, 10 μg/mL RBD, or 5 μg/mL LPS in the presence or absence of 1 mM D-cAMP for 2 h. The total cell lysates were subjected to Jess automatic western blot for phosphor- and total NF-κB. The blots were quantified based on intensity. (C and D) Soluble SARS2-S-induced GSDMD cleavage and inhibition by D-cAMP. HULEC-5a cells in suspension were treated as in (A and B) for 2 h. The total cell lysates were subjected to western blot for GSDMD. The cleavage of GSDMD was quantified based on intensity. (E) Soluble SARS2-S-induced release of IL-6 and IL-1β and the inhibition by D-cAMP. HULEC-5a cells in suspension were treated as in (A and B) for 2 h. The cell supernatants were subjected to ELISA for measuring the concentrations of IL-6 and IL-1β. Data are mean ± SD from three or four independent repeats. One-way ANOVA Tukey’s multiple comparisons test. *P < 0.05; **P < 0.01; and ****P < 0.0001.

Discussion

Whether integrins function as cell entry receptors for SARS-CoV-2 remains a topic of debate. Our argument is that SARS-CoV-2 cannot effectively infect cells in the absence of ACE2 by using integrins as receptors. Considering the widespread expression of integrins such as α₅β₁ and α_Vβ₃ in our body, if SARS-CoV-2 were able to infect ACE2-null cells using integrins as receptors, it could lead to more severe disease progression. It is worth noting that mouse and human integrins often share similar ligand-binding properties. However, mice without human ACE2 are not susceptible to SARS-CoV-2 infection (44–47). Our data based on pseudovirus and replication-competent rVSV-S along with α₅-KO did not support the involvement of α₅β₁ in the S-mediated cell entry. However, we did find that α₅β₁ contributes to S-mediated cell–cell fusion to some extent, which involves the participation of α₅ CT. The rVSV-S virus has served as a widely accepted model for the study of SARS2-S-mediated cell entry and inhibition, demonstrating a strong correlation with the authentic SARS-CoV-2 virus (48–50). However, distinctions between VSV and SARS-CoV-2 virions, such as variations in density, distribution, stability, and potential mutations of the S proteins on the viral envelope, can potentially impact the cell fusion function of the S protein. The limitation of using rVSV-S in our study was partially mitigated by using the S protein that expressed on cell surface to mimic the cell fusion between virus-infected cells, showing consistent results with rVSV-S. Nevertheless, we recognize the need for further comprehensive studies to compare our findings with those obtained from authentic SARS-CoV-2 virus.

The RGD-based interaction between integrins and SARS2-S is another subject of debate, with conflicting findings from different research groups. For example, one study reported a surprising high-affinity binding between SARS2-S S1 and α₅β₁ (12), whereas others found no binding at all (18, 19). Similarly, one study detected an interaction between SARS2-S RBD and α_IIbβ₃ (17), while another study showed no binding (18). Interestingly, a different study reported similar levels of binding between SARS2-S RBD and both RGD-recognizing (α₅β₁) and non-RGD-recognizing (α₄β₁, α₄β₇, and α_Lβ₂) integrins (14). The reasons for these inconsistencies among studies are not yet clear. Unlike the RGD motifs found in well-characterized integrin ligands, which are typically located in a flexible loop known as RGD finger (51, 52), such as in fibronectin (SI Appendix, Fig. S1A), the RGD motif in SARS2-S resides within a rigid α-helical structure that is unlikely to undergo conformational changes. This makes the backbone and sidechain orientation of the RGD unfavorable for integrin binding (Fig. 3B), as indicated by a molecular dynamics simulation study (53). Our data obtained from ELISA, pull-down, competition, mutagenesis, and functional assays did not support the RGD-dependent interaction between SARS2-S and α₅β₁. Instead, we found that the interaction may depend on the S2 subunit of SARS2-S. Furthermore, our detailed analysis of purified α₅β₁ suggested the involvement of α₅β₁ homo-oligomerization in binding to SARS2-S. Clearly, further structural studies are necessary to gain a more comprehensive understanding of the interaction between SARS2-S and α₅β₁, as well as other integrins.

We did not observe any inhibition effect of ATN-161 on the interaction between α₅β₁ and SARS2-S, nor on S-mediated virus infection and cell fusion. Our analysis of the cryo-EM structure of α₅β₁ in complex with Fn7-10 raised concerns about the mechanism by which ATN-161 works as an inhibitor in functional studies of α₅β₁. Despite being designed based on the synergistic integrin binding site on Fn9 (SI Appendix, Fig. S1A), the R to C mutation in ATN-161 may significantly reduce its binding affinity with α₅β₁. One study suggested that ATN-161 might bind to the N terminus of β₁, but no supporting data were provided (54). Another study proposed that it interacts with the RBD of SARS2-S and blocks ACE2 binding (55). Molecular docking and molecular dynamics simulation studies have also suggested the binding of ATN-161 on the main protease of SARS-CoV-2 (56). Thus, the exact mechanism by which ATN-161 interacts with α₅β₁ remains unknown. Caution should be taken when considering the use of ATN-161 as a potential α₅β₁ antagonist.

SARS2-S has been reported to induce an inflammatory response via TLRs in macrophages. However, there are inconsistent and conflicting data regarding whether TLRs, or which TLRs (TLR2 or TLR4), interact with the SARS2-S (57–60). Integrin α₅β₁ has been implicated in regulating inflammatory response through its interaction with physiological or pathogen ligands (38, 61, 62). We used HUVEC as a model given the established inflammatory role of α₅β₁ in this cell line (38) and the absence of surface TLR2 and TLR4 expression in the resting state (63). We found that both immobilized and soluble SARS2-S induced an inflammatory response in endothelial cells, as demonstrated by NF-κB activation and cytokine release. This finding is consistent with another study (22). In agreement with our protein interaction results, the α₅β₁-dependent inflammatory responses were induced by the SARS2-S or its S2 subunit but not its RBD or S1 subunit. In contrast, SARS-S did not induce NF-κB activation or IL-6 release under our experimental conditions. This mechanism differs from the inflammation mediated by TLRs, which do not discriminate between SARS-CoV and SARS-CoV-2 (8, 64). It is conceivable that SARS-CoV-2 induces inflammation through processes not previously observed in SARS-CoV.

The involvement of PDE4D activation in α₅β₁-mediated endothelial inflammatory response has been established by both in vitro and in vivo studies (38, 65, 66). Upon activation, PDE4D hydrolyses cAMP, leading to reduced protein kinase A (PKA) activity (40). The reduction in PKA activity subsequently triggers NF-κB activation and the production of proinflammatory cytokines such as IL-6. Additionally, decreased intracellular cAMP levels activate the NLRP3 inflammasome, resulting in the maturation and secretion of IL-1β (41). Our findings that SARS2-S, via α₅ integrin, induces GSDMD cleavage and IL-1β release, a downstream response of inflammasome activation (43), align with this scenario. Previous studies have demonstrated the ability of SARS2-S to induce NLRP3 inflammasome activation in macrophages (67, 68). Besides endothelial cells, α₅ integrin and PDE4D are widely expressed in various other cell types, including immune cells such as macrophages and neutrophils, lung epithelial cells, and neurons. These cell types are known to play roles in COVID-19 pathogenesis (2, 5). Elevated levels of IL-6 and IL-1β and inflammasome activation have been associated with the severity of COVID-19 (3, 69).

Our data support the involvement of the inflammasome and GSDMD in COVID-19 (4), mediated by the SARS2-S-induced α₅-PDE4D-cAMP cascade. PDE4 inhibition has been validated as an anti-inflammatory strategy and may hold potential as a treatment for COVID-19 (70, 71). While our data suggest the important role of cAMP regulation by the SARS2-S/α₅β₁ signaling in the inflammatory response, further investigation into the involvement of additional α₅β₁-associated signaling pathways is required. It also remains to be determined whether the SARS2-S/α₅β₁ interaction and signaling follow the conventional integrin activation model that involves large-scale conformational regulation. Our study provides a molecular basis for targeting the S-α₅β₁ interaction and its downstream pathway as potential therapeutic approaches for COVID-19.

Materials and Methods

The detailed materials and methods are available in SI Appendix. DNA constructs were either obtained from Addgene or generated through standard molecular cloning techniques. We created α₅ integrin knockout cell lines using CRISPR/Cas9 technology. Recombinant proteins were either obtained from BEI Resources or produced in HEK293 cells. Virus infection and cell–cell fusion assays were conducted following established protocols. The interaction between the S proteins and α₅β₁ was assessed through ELISA, DLS, and pull-down assays using standard procedures. NF-κB activation, cytokine release, and GSDMD cleavage were analyzed using ELISA and western blotting. Statistical analysis was carried out on at least three individual datasets and analyzed with GraphPad Prism software.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix. For additional information and resource requests, please contact Dr. Jieqing Zhu at Jieqing.Zhu@versiti.org.