AXL Promotes SARS-CoV-2 Infection of Pulmonary and Bronchial Epithelial Cells

The current coronavirus disease 2019 (COVID-19) pandemic presents a global public health challenge. The viral pathogen responsible, Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), binds to a host receptor ACE2 through its spike (S) glycoprotein, which mediates membrane fusion and virus entry. Although the role of ACE2 as a receptor for SARS-CoV-2 is clear, studies have shown that ACE2 expression across different human tissues is extremely low, especially in pulmonary and bronchial cells. Thus, other host receptors and/or co-receptors that promote the entry of SARS-CoV-2 into cells of the respiratory system might exist. In this study, we have identi�ed tyrosine-protein kinase receptor UFO (AXL), speci�cally interacts with SARS-CoV-2 S on the host cell membrane. When overexpressed in cells that do not highly express either AXL or ACE2, AXL promotes virus entry as e�ciently as ACE2. Strikingly, deleting AXL, but not ACE2, signi�cantly reduces infection of pulmonary cells by the SARS-CoV-2 virus pseudotype. Soluble human recombinant AXL, but not ACE2, blocks SARS-CoV-2 virus pseudotype infection in pulmonary cells. Taken together, our �ndings suggest AXL may play an important role in promoting SARS-CoV-2 infection of the human respiratory system and is a potential target in future clinical intervention strategies.


Main Text
COVID-19 has caused a global pandemic since December 2019 and presents a global public health threat.The viral pathogen responsible, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is a highly contagious enveloped, positive-strand RNA virus 1 that causes upper respiratory diseases, fever and severe pneumonia in humans 1,2 .SARS-CoV-2 belongs to the β coronavirus genus.Other family members include severe acute respiratory syndrome coronavirus (SARS-CoV) and middle east respiratory syndrome coronavirus (MERS-CoV), which caused outbreaks in 2003 and 2012, respectively, though on a much smaller scale 3,4 .SARS-CoV-2 preferentially infects cells of the respiratory tract 5 but has also been detected in almost all human organs, including the lungs, pharynx, heart, liver, brain, kidneys and digestive system [6][7][8] .
Coronavirus binds to host receptors through its spike (S) glycoprotein, which mediates membrane fusion and virus entry 9 .The S protein is cleaved into the N-terminal S1 subunit and C-terminal S2 subunit by host proteases Transmembrane protease serine 2 (TMPRSS2) and FURIN (Fig. 1a) 10 .SARS-CoV-2 shares 79.5% genetic identity with SARS-CoV, and studies have shown that ACE2, a cellular receptor for SARS-CoV 11 , also binds SARS-CoV-2 S and serves as the entry point for SARS-CoV-2 (Fig. 1a) 1,12 .The complex structures between ACE2 and the receptor binding domain (RBD) of SARS-CoV-2 S have recently been resolved 13,14 .Although the role of ACE2 as the receptor for SARS-CoV-2 is quite clear, the e cacy of drugs targeting ACE2 is still debatable.Current efforts regarding therapeutic interventions are focused on developing monoclonal antibodies targeting the SARS-CoV-2 S protein [15][16][17] .However, no SARS-CoV monoclonal antibodies isolated to date are able to neutralize SARS-CoV-2 18,19 , highlighting the importance of understanding the infection machinery in detail.
Although SARS-CoV-2 infection largely manifests as respiratory system symptoms, single-cell sequencing data indicate that overall ACE2 expression across different human tissues is low, especially in pulmonary and bronchial cells 20,21 .In addition, a recently published single-cell mRNA sequencing dataset of 232,905 single cells from major adult organs 22 revealed that ACE2 is speci cally expressed in the kidney but rarely in organs such as the lung and trachea.Thus, other important host receptors and/or coreceptors that facilitate the entry of SARS-CoV-2 into cells of the respiratory system might exist.
Using TAP-MS to analyse protein complexes interacting with SARS-CoV-2 S in pulmonary and bronchial cells, we identi ed 22 candidates that may facilitate SARS-CoV-2 entry.The candidates were further enriched via molecular simulations and then tested through a series of biochemical and cell biology assays.We found that AXL speci cally interacts with SARS-CoV-2 S on the host cell membrane and facilitates virus entry.Overexpressing AXL in HEK293T cells promotes virus entry as e ciently as ACE2.Downregulating AXL, but not ACE2, signi cantly reduces infection of pulmonary cells by the SARS-CoV-2 virus pseudotype.Soluble human recombinant AXL, but not ACE2, blocks SARS-CoV-2 virus pseudotype infection in pulmonary cells.Taken together, our results suggest that AXL promotes SARS-CoV-2 entry into human pulmonary and bronchial epithelial cells.
To identify host proteins responsible for SARS-CoV-2 infection of pulmonary and bronchial cells, we explored those interacting with SARS-CoV-2 S in A549, H1299 and BEAS-2B cells.To this end, we overexpressed FLAG-tagged full-length SARS-CoV-2 S, SARS-CoV-2 S RBD, or SARS-CoV-2 S S1+S2 in these cells or added recombinant full-length FLAG-tagged SARS-CoV-2 S or SARS-CoV-2 S RBD to the cells.Only cells overexpressing full-length SARS-CoV-2 S showed strong membrane localization of the S protein (Extended Data Fig. 2a).Moreover, the full-length SARS-CoV-2 S was successfully cleaved in A549, H1299 and BEAS-2B cells, generating fragments with sizes similar to S1-cleaved SARS-CoV-2 S (Extended Data Fig. 2b).Thus, we established A549, H1299 and BEAS-2B cells stably expressing SARS-CoV-2 S or control in uenza A virus (A/Guangzhou/39715/2014 (H5N6)) haemagglutinin (HA), which were both fused to Nterminal SFB triple tags (S tag-2x Flag tag-SBP tag) (Fig. 2).Twelve clones of each bait in each cell line were examined in follow-up experiments, and bait protein expression and localization were con rmed using Western blotting and immunostaining, respectively.We chose two clones of each bait in each cell line with membrane/cytosol localization and moderate expression as biological repeats for TAP-MS analysis and isolated, combined, and a nity-puri ed membrane and soluble fractions using TAP.
Proteins associated with the bait in isolated complexes were identi ed by MS, and human, SARS-CoV-2 and H5N6 databases were searched for these proteins (Fig. 2, Supplementary Tables 2, 3).TAP-MS recovered a reasonable number of peptides, with high coverage for the bait proteins (Extended Data Fig. 2c, Supplementary Table 2) and reasonable data reproducibility (Extended Data Fig. 2d).In total, we identi ed 59,895 peptides of 9,479 proteins, representing 3,725 unique preys (Supplementary Table 3), 524 of which are membrane proteins or are able to translocate to the membrane.Interestingly, we did not identify ACE2 in any of our experiments, indicating ACE2 may not be the major host receptor of SARS-CoV-2 in these cells due to its low expression (Extended Data Fig. 2e).Nonetheless, we did consistently identify ACE2 in TAP-MS for SARS-CoV-2 S in HEK293T cells, which express very low levels of ACE2 1 , indicating that our approach is able to recover host receptors for SARS-CoV-2 S (Extended Data Fig. 2e, Supplementary Table 3).
To distinguish bona de SARS-CoV-2 S-interacting proteins from the large number of non-speci c interactors frequently obtained in AP-MS results 23 , we carried out H5N6 HA TAP-MS experiments under identical experimental conditions as controls and assigned a quality-associated probabilistic score to each binary interaction using the MUSE algorithm 24,25 to remove non-speci c interacting proteins.Twenty-two high-con dence candidate receptors/co-receptors that were present in protein complexes identi ed from at least two cell lines were chosen.

Computational screening of the top candidate receptors for SARS-CoV-2 S
To nd host receptor(s) with the highest binding a nity for SARS-CoV-2 S, TAP-MS results were further enriched using a hierarchical computational protocol combining protein-protein docking, molecular modelling, MD simulations, and MM/PBSA calculations (Fig. 2).For each candidate receptor, we docked its extracellular domain with the full-length SARS-CoV-2 S and performed multiple rounds of conformational optimization using HADDOCK 26 .Top-ranking poses were ltered by building full-length proteins and excluding complex conformations with potential steric clashes.Moreover, the stability of these docking poses was further assessed with explicit solvent atomistic simulations.Elevennanosecond MD simulations were performed for each protein-protein docking conformation; to estimate the most likely binding conformation and the corresponding binding a nity, MM/PBSA calculations were conducted using the last 1-ns MD trajectories.Although MM/PBSA is not able to provide quantitatively accurate binding a nities, we consider the computation to be an approach for prioritizing candidate receptors for viral entry experiments.The top three candidate receptors with the most favourable a nity scores, namely, AXL, epidermal growth factor receptor (EGFR), and low-density lipoprotein receptor (LDLR) (Fig. 2), were subjected to further biochemical experiments.

SARS-CoV-2 S interacts with host AXL
We rst validated the observed interactions using FLAG-tagged SARS-CoV-2 S and MYC-tagged receptor proteins.All three top candidates exhibited a binding a nity similar to that of ACE2 by SARS-CoV-2 S in vivo (Fig. 3a).AXL and SARS-CoV-2 S were mainly co-localized to the cell membrane, as is ACE2, whereas LDLR and EGFR were not (Fig. 3b).In addition, recombinant His-tagged AXL was able to bind FLAGtagged SARS-CoV-2 S puri ed from HEK293T cells in vitro (Fig. 3c).Furthermore, recombinant mFctagged SARS-CoV-2 S pulled down AXL in H1299 cells, indicating that extracellular S is also capable of binding to AXL on the cell surface (Supplementary Table 4).To look into the interaction between AXL and SARS-CoV-2 S, the interface was analysed and an extensive hydrophilic network was found along the interface.In contrast to ACE2 receptor, AXL interacted with spike NTD domain, rather than RBD domain (Extended Data Fig. 3).
Next, we measured the protein and mRNA levels of these proteins in HEK293T, A549, H1299 and BEAS-2B cells.The level of ACE2 expression was low in all the cell lines tested; LDLR and EGFR expression was detected in all cell lines, and AXL was highly expressed in A549, H1299 and BEAS-2B cells (Figs. 3d, e).
AXL belongs to the phosphatidylserine receptor family, many of which may serve as factors or co-factors for viral entry 27 .Regardless, none of the other receptors in this family were speci cally recovered by SARS-CoV-2 S TAP-MS or AP-MS in A549, H1299 or BEAS-2B cells (Supplementary Tables 2-4), indicating that this interaction between AXL and SARS-CoV-2 S is highly speci c.AXL is a receptor tyrosine kinase that transduces signals from the extracellular matrix into the cytoplasm 28 and regulates many physiological processes, including cell survival, proliferation, differentiation and immunity [29][30][31][32] .AXL forms a stable complex with its ligand growth arrest-speci c protein 6 (GAS6) via interaction interfaces involving its Ig-like domains 33 .In fact, AXL has been reported to act as a cell surface receptor for Ebola virus 34 and plays important roles in mediating virus-host interactions 27 .
Analysis of the human cell landscape at the single-cell level revealed that AXL is expressed in most human tissues tested (Extended Data Fig. 1c, Supplementary Table 1).For example, many more cells in the lung and trachea expressed AXL (1232/17628 and 456/9521, respectively) than expressed ACE2 (16/17628 and 20/9521, respectively) (Figs.3f-i, Supplementary Table 1).Thus, AXL might play important roles in facilitating SARS-CoV-2 entry into pulmonary and bronchial epithelial cells.

AXL facilitates SARS-CoV-2 entry into human cells
To determine whether SARS-CoV-2 uses AXL to facilitate entry into human cells, we conducted virus infectivity analyses using the SARS-CoV-2 virus pseudotype system in HEK293T cells, which express very low levels of ACE2 and AXL (Figs. 3d, e) and cannot be infected by the SARS-CoV-2 virus pseudotype (Fig. 1h).We overexpressed ACE2, AXL, LDLR, EGFR or the empty vector in HEK293T cells and infected the cells with the SARS-CoV-2 virus pseudotype.Viral infection of HEK293T cells was signi cantly enhanced in those overexpressing ACE2, reproducing previous results 1,12 and con rming that ACE2 is a SARS-CoV-2 receptor.Similar enhancement was observed for viral infection of HEK293T cells overexpressing AXL but not cells expressing LDLR or EGFR (Figs. 4a, b).Moreover, SARS-CoV-2 virus pseudotype particles were only observed in cells overexpressing AXL (Fig. 4c).We also overexpressed tyrosine-protein kinase Mer (MER) or broblast growth factor receptor (FGFR), which belong to the same TAM/TAM-like receptor families of receptor tyrosine kinases.However, MER and FGFR both failed to promote viral entry into cells, also indicating that the function of AXL in facilitating viral entry is highly speci c (Figs. 4d-f).To identify the region of AXL that binds the SARS-CoV-2 S protein, we generated several truncation mutants for AXL and found its extracellular N-terminal domain, but not the kinase domain, to be responsible for the interaction (Fig. 4g) and viral entry into host cells (Figs. 4h, i).In addition, our results show that SARS-CoV-2 pseudoviral particles colocalized with endocytosis and vesicle tra cking markers including Early endosome antigen 1(EEA1), DCC-interacting protein 13-alpha (APPL1), Clathrin heavy chain 1 (CLTC) and Syntaxin-6 (STX6), indicating AXL-facilitated viral entry utilizes the host endocytosis system (Fig. 4j).
Taken together, these results indicate that AXL facilitates SARS-CoV-2 entry as potently as ACE2.

AXL is required for SARS-CoV-2 entry into human pulmonary epithelial cells
To evaluate the signi cance of AXL in SARS-CoV-2 infection of pulmonary epithelial cells, we knocked down ACE2, AXL, LDLR, and EGFR (Extended Data Fig. 4) in H1299 and A549 cells by transfecting corresponding siRNAs and then infected the cells with the SARS-CoV-2 virus pseudotype (Figs.5a, b).Downregulation of AXL, but not ACE2, LDLR or EGFR, drastically reduced SARS-CoV-2 virus pseudotype infection of H1299 and A549 cells at 24 h post-infection (Figs.5a-d), indicating that AXL is required for SARS-CoV-2 entry into these cells.Next, we established H1299 AXL-KO or ACE2-KO cells using the CRISPR-Cas9 system (Fig. 5e).Knocking out AXL but not ACE2 completely prevented viral infection in H1299 cells (Figs. 5f, g), indicating that AXL is required for SARS-CoV-2 entry into pulmonary and bronchial epithelial cells.
Soluble human recombinant ACE2 has reduced SARS-CoV-2 recovery from Vero cells which expresses high level ACE2 35 .To further con rm AXL's indispensable role in SARS-CoV-2 infections of pulmonary and bronchial cells, and determine the potential impact on clinical practice, we added soluble human recombinant ACE2 or AXL into HEK293T cells overexpressing ACE2 or AXL, and then infected the cells with the SARS-CoV-2 virus pseudotype (Figs.6a-d).Soluble human ACE2 and AXL blocks SARS-CoV-2 virus pseudotype infection in cells overexpressing ACE2 and AXL, respectively.However, they failed to block viral infection in cells overexpressing each other (Figs.6a-d), indicating AXL's function in mediating viral entry is very likely to be independent of ACE2.Soluble human AXL but not ACE2 blocks SARS-CoV-2 virus pseudotype infection in H1299 cells (Figs. 6e-g), con rming that AXL is required for SARS-CoV-2 entry into pulmonary epithelial cells.
Taken together, our results not only con rm that the human tyrosine-protein kinase receptor AXL speci cally interacts with the SARS-CoV-2 S glycoprotein but also clarify its indispensable role in facilitating SARS-CoV-2 entry into human pulmonary epithelial cells, and the cellular signi cance deserves further investigation (Fig. 6h).

Discussion
COVID-19 primarily causes respiratory system illness; however, an increasing number of case reports of SARS-CoV-2 viral infection have shown that it can also affect almost all of the body's primary organs, including the lungs, pharynx, heart, liver, brain, and kidneys 7 .SARS-CoV-2 viral particles were originally visualized in ultrathin sections of human airway epithelial cells 1 .Later, SARS-CoV-2 RNA was detected in oesophagus, stomach, duodenum, and rectum specimens 8 as well as in the kidney 7 .These results are consistent with single-cell sequencing data (Supplementary Table 1) showing that ACE2 is highly expressed in the kidney and digestive system.However, in other tissues, including the heart (0.3%), liver (0.1%), brain (0%), lung (0.1%) and trachea (0.2%), ACE2 expression was under 1% (Supplementary Table 1).Accordingly, we speculated on the existence of additional proteins that are responsible for viral entry in these tissues.
In our current study, we found that SARS-CoV-2 is able to utilize either ACE2 or AXL for entry into human cells.AXL is widely expressed in almost all human organs, and particularly in human pulmonary and bronchial epithelial tissue and cells, its expression is much higher than that of ACE2 (Figs. 1b-g, 3g-j, Extended Data Fig. 1, Supplementary Table 1).Given that AXL does not co-express with ACE2 or TMPRSS2 in the human lung or trachea (Extended Data Fig. 5, Supplementary Table 1), and the fact they could not block viral infection when the other protein is overexpressed (Figs.6a-d), AXL's function in mediating SARS-CoV-2 infection is very likely to be independent of ACE2.Nevertheless, further investigation is needed to determine whether AXL and ACE2 utilize the same co-factors or are involved in similar infection processes.
Since the outbreak of COVID-19, extensive effort has been devoted to developing drugs that target the human cell viral receptor ACE2.Unfortunately, the outcomes were not satisfying.Considering the low expression level of ACE2 in human pulmonary and bronchial epithelial cells, downregulation of ACE2 elicited a minor effect in reducing SARS-CoV-2 virus pseudotype infection of H1299 and A549 cells (Figs. 5a-d), indicating that ACE2 might not be a good drug target for the lungs and bronchi.
As AXL is overexpressed in numerous cancer types, it has already been pursued as a drug target.
However, we note that most reported AXL inhibitors target its kinase domain and would thus not have e cacy against SARS-CoV-2 infection because the virus interacts with the extracellular Ig-like domains of AXL (Figs. 4 g-i).Alternative approaches include targeting the AXL extracellular Ig-like domains and the region of the SARS-CoV-2 S protein that interacts with AXL.Clinical grade human recombinant soluble AXL may be used to block viral infection (Figs. 6 c-g).Based on our data, AXL is likely to bind the Nterminal region of SARS-CoV-2 S, unlike the binding of ACE2 to the RBD (Extended Data Fig. 3).Interestingly, a recent study identi ed a potent neutralizing human antibody that binds to the N-terminal domain of SARS-CoV-2 S 36 .Although this antibody does not bind to the RBD or block its interaction with ACE2, its neutralizing activity was much higher than that of antibodies targeting the RBD 36 .This report highlights the importance of the N-terminal domain of SARS-CoV-2 S during viral infection and may support the important role of AXL during infection of the human pulmonary and bronchial systems.
We anticipate that upon further validation and mechanistic understanding, our ndings will contribute to the development of therapeutic solutions to COVID-19.
Tandem a nity puri cation H1299, A549, BEAS-2B and HEK293T cells stably expressing an N-terminal SFB-fused SARS-CoV-2 S protein or control in uenza A virus haemagglutinin were selected via culture in medium containing 2 μg/ml puromycin.Protein expression was con rmed by immunostaining and Western blotting as described previously 37 .For TAP, the membrane and soluble proteins of 1 × 10 8 cells were extracted using a membrane/soluble protein isolation kit (Beyotime, China) with protease inhibitors at 4°C.The lysates were combined and incubated with streptavidin-conjugated beads (Thermo Fisher Scienti c, USA) for 2 h at 4°C.The beads were washed three times with 1 × NETN buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40), and the bound proteins were eluted with NETN buffer containing 2 mg/ml biotin (Sigma-Aldrich, USA) for 2 h at 4°C.The eluates were incubated with S-protein beads (Millipore, USA) for 1 h; the beads were washed three times with NETN buffer and subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis.Each pull-down sample was electrophoresed.The whole band was excised as one sample and subjected to in-gel trypsin digestion and LC-MS.
Liquid chromatography and mass spectrometry LC and MS were performed as described previously 24,25 .Brie y, the gel bands excised as described above were cut into approximately 1-mm 3 pieces, which were then subjected to in-gel trypsin digestion 38 and dried.The samples were reconstituted in 5 ml of high-performance liquid chromatography solvent A (2.5% acetonitrile and 0.1% formic acid).A nanoscale reverse-phase high-performance liquid chromatography capillary column was created by packing 5-mm C18 spherical silica beads into a fused silica capillary (100 mm inner diameter × ~20 cm length) using a ame-drawn tip.After the column was equilibrated, each sample was loaded using an autosampler.A gradient was formed, and peptides were eluted with increasing concentrations of solvent B (97.5% acetonitrile and 0.1% formic acid); the peptides were subjected to MS as they eluted.
The MS conditions were as follows.For Orbitrap Fusion Lumos, the source was operated at 1.9 kV, with no sheath gas ow and with the ion transfer tube at 350°C.The mass spectrometer was programmed to acquire in a data-dependent mode.The survey scan was from m/z 350 to 1500, with a resolution of 60,000 at m/z 200.The 20 most intense peaks with charge state 2 and greater were acquired with collision-induced dissociation with a normalized collision energy of 30% and one microscan; the intensity threshold was set at 1000.MS2 spectra were acquired with 15,000 resolution.The peptides were detected, isolated, and fragmented to produce a tandem mass spectrum of speci c fragment ions for each peptide.
Mass spectrometry data analysis MS peptide sequences and, hence, protein identity were determined by matching fragmentation patterns in protein databases using the Mascot software program (Matrix Science, USA).Enzyme speci city was set to partially tryptic with two missed cleavages.Modi cations of the peptides included carboxyamidomethylation (cysteines, variable), oxidation (methionine, variable), phosphorylation (S, T, Y, H, variable) and acetylation (N-term, K, variable).Mass tolerance was set to 20 ppm for precursor ions and fragment ions.UniProt was the database searched (Homo sapiens, SARS-CoV-2 and In uenza A virus (A/Guangzhou/39715/2014(H5N6))).Spectral matches were ltered for a false-discovery rate of less than 1% at the peptide level using the target-decoy method 39 , and protein inference was considered following the general rules 40 , with manual annotation applied when necessary.This same principle was used for protein isoforms when they were present; in general, the longest isoform is reported.Interaction ltration was further performed using the MUSE algorithm as described previously 24,25 to assign quality scores for the identi ed PPIs.TAP-MS performed under identical experimental conditions for in uenza A virus haemagglutinin from H1299 cells was used as a true negative control for the MUSE analysis.
Enrichment of the overall and individual groups of proteins in canonical signalling pathways, functional categories and diseases was assessed.P values were estimated using Knowledge Base included with the Ingenuity Pathway Analysis software program (Ingenuity Systems, USA).Only statistically signi cant correlations (P < 0.01) are shown.The -log (P value) for each function and related HCIPs is listed.

Analysis of protein expression levels in human tissue cells
Analysis of ACE2 and AXL expression in the cells of human tissues was performed using a recently published single-cell mRNA sequencing dataset consisting of 232,905 single cells from all major adult organs after the batch gene background removed and cells with detection less than 500 transcripts ltered out.For digital gene expression data matrices were ln(CPM/100+1) transformed and downstream procedures for ltering and dimensional reduction were performed using Seurat v3.1.0 41.All the genes were used for initial principle component analysis, and the top 10 principal components were used for nonlinear dimensional reduction (t-SNE) analysis.

Computational modelling of protein-protein complexes
Protein-protein docking calculations were performed using HADDOCK 26 .For each receptor, 600,000 docking poses were randomly generated and rigid-body minimized in the rst stage, after which 400 poses with the lowest HADDOCK scores were selected for optimization of the protein-protein interface with simulated annealing.After further re nement of these poses in an explicit 8-Å water layer environment, clustering analysis was conducted, and the lowest score poses for each of the top 20 largest clusters were selected.One additional lter was performed by building the atomistic model of the full-length receptors interacting with SARS-CoV-2 S and excluding any conformation leading to potential steric clashes.
In the next stage, molecular dynamics simulations were carried out to assess the stability of these docking poses.Simulation systems were established using CHARMM 42 with selected complexes solvated in a periodic 210×210×210-Å 3 cubic TIP3P water box and neutralized with extra K + or Cl -ions.
The systems were subjected to 11-ns NPT simulations with the CHARMM36 m protein force eld 43 using OpenMM 44 .Non-bonded interactions were truncated at 12 Å with smooth switching from 10 Å, and electrostatic interactions were calculated using the particle meshed Ewald (PME) method.MD trajectories evolved at 300 K with a 2-fs timestep in which bonds involving H atoms were constrained.To provide a rough estimation of the binding a nity between the candidate receptors and full-length SARS-CoV-2 S, MM/PBSA calculations were carried out using the last 1-ns MD trajectories.

Western blotting and immunoprecipitation
Cells were harvested and lysed in NETN buffer on ice for 30 min.After measuring the protein concentration using a BCA kit (Thermo Fisher Scienti c, USA), 5× loading buffer (Beyotime, China) was added, and the mixture was boiled for 15 min.After 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis, the proteins were transferred onto PVDF membranes (Millipore, USA).The membranes were incubated with the indicated primary antibodies at 4°C overnight, washed three times with PBS-T buffer (1 × PBS, 0.05% V/V TWEEN 20) for a total of 15 min and incubated with secondary antibodies (1:2000, Cell Signaling Technology, CST, USA) for 1 h at room temperature.Signals were detected using an ECL kit (Pierce, USA).
The following primary antibodies were used: anti-MYC For IP and co-IP assays, 1 × 10 7 cells were lysed in NETN buffer on ice for 30 min.The cell lysates were precleared by incubation with 5 μl of magnetic beads (Millipore, USA) for 1 h and then with the indicated antibodies at 4°C with rotation overnight.After centrifugation, the supernatant was incubated with 10 μl of magnetic beads at 4°C with rotation for 1 h.The beads were washed three times with cold NETN buffer using a magnetic separator (Millipore, USA), followed by elution with 40 μl of protein lysis buffer (Beyotime, China).Ten microlitres of 5 × loading buffer (Beyotime, China) was added, and the mixture was boiled for 15 min and subjected to Western blotting.
The cells were viewed using an Olympus IX73 Microscope Imaging System (Olympus, Japan).

SARS-CoV-2 virus pseudotype production and infection
The SARS-CoV-2 virus pseudotype was packaged and used to infect cells as previously described 45 .
Brie y, HEK293T cells were cultured in 10-cm plates pre-coated with poly-l-lysine and incubated with DMEM supplemented with 10% FBS, penicillin/streptomycin and l-glutamine.The next day, the cells were co-transfected with psPAX2, pLenti-GFP, and SARS-CoV-2 S plasmids using Jetprime DNA transfection reagents (Polyplus, France) according to the manufacturer's instructions.Supernatants were collected at 48 and 72 h post-transfection, mixed with PEG overnight, passed through a 0.45-μm lter, centrifuged at 500× g for 5 min, aliquoted and stored at -80 °C.
To transduce cells with the SARS-CoV-2 virus pseudotype, HEK293T, H1299 or A549 cells were seeded into 24-well plates, transfected with the indicated plasmids or siRNAs overnight, and then infected with the SARS-CoV-2 virus pseudotype for 24 h.Cells were washed 3 times and viewed using an Olympus IX73 Microscope Imaging System (Olympus, Japan).To evaluate SARS-CoV-2 virus pseudotype infection titration, H1299 cells were cultured in 24-well plates.25-200 µg/ml human recombinant HIS-AXL (RP-HIS-AXL) or HIS-ACE2 (RP-HIS-ACE2) were mixed with SARS-CoV-2 virus pseudotype (10 7 PFUs, MOI 5) for 30 min and then added to the culture medium of H1299 cells.Cells were washed after 2 h post-infection and incubated with fresh medium.Cell were recovered 24 hr, and viral RNA was assayed by RT-qPCR.

Declarations
Data analysis and ethics statement Statistical signi cance between two groups was determined by unpaired two-tailed Student's t test.
Differences were considered to be signi cant at p < 0.05.This study was approved by the Ethics Committee of Westlake University.
Figures   AXL expression levels in (f, g) pulmonary and (h, i) bronchial cells were evaluated using the human cell landscape at the single-cell level.Gene expression for each cell type was visualized using (f, h) tSNE and (g, i) violin plots.

Figure 3 The
Figure 3