A pH-driven receptor switch from ACE2 to NPC1 enables endosomal entry of SARS-CoV-2

TL;DR

The study shows that SARS-CoV-2 switches from using ACE2 to NPC1 as a receptor during endosomal entry, depending on pH levels.

Contribution

The novel contribution is the discovery of a pH-dependent receptor switch from ACE2 to NPC1 during SARS-CoV-2 endosomal entry.

Findings

NPC1 is an essential intracellular receptor for spike-mediated fusion in acidic endosomes.

Omicron variant shows increased dependence on NPC1 for endosomal entry.

NPC1 binds with higher affinity to the spike protein under acidic conditions compared to ACE2.

Abstract

SARS-CoV-2 enters cells either by TMPRSS2-mediated fusion at the cell surface or by cathepsin-dependent fusion from late endosomes, yet how spike remains competent for fusion during endocytic trafficking is unclear. Here we identify Niemann–Pick C1 (NPC1) as an essential intracellular receptor that enables spike-mediated fusion in acidic late endosomes/lysosomes. Genetic ablation or pharmacological inhibition of NPC1 selectively blocked endosomal entry but not the surface entry, and Omicron showed heightened NPC1 dependence consistent with reduced fusogenicity and preferential endosomal entry in cell culture. A bimolecular fluorescence complementation assay localized spike–NPC1 complexes to Rab7/LAMP1-positive compartments. Biochemical and competition analyses, supported by structural modelling, revealed that NPC1 loop C engages a surface on the receptor-binding domain distinct from the ACE2 interface and binds with higher affinity under acidic conditions, whereas ACE2 binding is favored at neutral pH. These findings define a pH-driven receptor switch from ACE2 to NPC1 during endocytic entry.

Figures

F](#F1)) and WB ([Fig. 1G](#F1)). As controls, 293T and A549 cells were transduced with ACE2 (A) or ACE2 and TMPRSS2 (AT). Endogenous ACE2 expression was low in all parental lines, but high ACE2 expression was detected in transduced 293T-A, 293T-AT, A549-A, and A549-AT cells. Despite low ACE2 expression, Vero, Caco-2, and Calu-3 supported robust infection, indicating that high ACE2 levels are not essential for SARS2 infection. Endogenous TMPRSS2 expression, however, was high in Calu-3 and Caco-2, but low in A549 and Vero, and high TMPRSS2 expression was detected in transduced cells. Thus, cell

C](#F2)-[D](#F2)). *NPC1*-KO also impaired SARS2-S PV infection in A549-A and Vero-A cells, whereas this effect was substantially attenuated in A549-AT and Vero-AT cells. The same pattern was observed with authentic SARS2 infection ([Fig. 2E](#F2)). Notably, residual SARS2 infection persisted in *NPC1*-KO A549-A and Vero-A cells, particularly at 48 h post-infection ([Fig. 2E](#F2)), likely reflecting low-level surface entry mediated by endogenous TMPRSS2 or related proteases.

C](#F3)). *NPC1*-KO significantly reduced Delta and Gamma PV infection in ACE2-only cells but had little effect in TMPRSS2-expressing cells. In striking contrast, Omicron PV infection was strongly suppressed by *NPC1*-KO even in TMPRSS2-expressing cells. Similarly, the NPC1 inhibitor Tub-III selectively inhibited Delta and Gamma PV infection in A549-A cells but not in A549-AT cells, while robustly suppressing Omicron PV infection irrespective of TMPRSS2 expression ([Fig. 3D](#F3)). Authentic Delta and Omicron virus infections yielded comparable results ([Fig. 3E](#F3)).

C](#F5)), indicating a putative NPC1-C binding site located between the epitopes of B1-182.1 and A19-46.1 and distinct from the ACE-binding site ([Fig. 5D](#F5), left). This interaction pattern is consistent with the model predicted by AlphaFold3, in which NPC1-C engages a region on RBD overlapping with the epitopes of both class I and class II antibodies but away from the ACE2-binding site ([Fig. 5D](#F5), right).

B](#F6)-[C](#F6), Fig. S7). EPM and EPM-1C exhibited intermediate activity, NPC1L1-1C supported modest fusion, whereas NPC1, NPC1L1, and EPM-L1C showed no detectable activity. Notably, reducing the pH to 5.0 had no effect on EPM-mediated fusion but markedly decreased ACE2-mediated fusion and, importantly, significantly enhanced EPM-1C–mediated fusion ([Fig. 6D](#F6)).

Full text

Introduction

Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) continues to circulate globally and cause significant disease, yet key aspects of its fundamental entry biology remain unresolved. SARS-CoV-2 initiates infection when the viral spike (S) glycoprotein engages its well-established cell-surface receptor, angiotensin-converting enzyme 2 (ACE2) ^1^. The S protein, a type I integral membrane protein, is cleaved during virus assembly by furin into two functional subunits: S_1_, which contains the receptor-binding domain (RBD), and S_2_, which encodes elements required for membrane fusion. This furin-cleavage is inhibited by the host MARCH E3 ubiquitin ligases ^2, 3^. Upon ACE2 binding, S_1_ dissociates from S_2_, triggering conformational rearrangements that prime S_2_ for fusion.

SARS-CoV-2 can enter cells through two mechanistically distinct pathways. At the plasma membrane, transmembrane serine protease 2 (TMPRSS2) cleaves S_2_ at the S_2_' site to expose the fusion peptide, enabling immediate cell-surface fusion. In cells lacking TMPRSS2 activity, virions instead undergo endocytosis, where cathepsin B/L cleavage within acidified late endosomes activates S_2_ and promotes endosomal fusion.

Niemann–Pick C1 (NPC1) is a 1,278-amino-acid membrane protein containing 13 transmembrane helices and three major luminal loops (A, C, and I). Localized to late endosomes, NPC1 cooperates with NPC2 to export cholesterol ^4^, but also serves as an essential entry receptor for filoviruses ^5, 6^. Notably, three independent genome-wide CRISPR screens have identified NPC1 as a critical host factor for SARS-CoV-2 infection ^7, 8, 9^. Consistent with this, a natural-compound screen from our group identified Tubeimosides as inhibitors of Ebola virus (EBOV) glycoprotein (GP) binding to NPC1 loop C (NPC1-C), and unexpectedly, these compounds also potently blocked SARS-CoV-2 entry by targeting NPC1 ^10^.

Here, we define distinct and complementary roles for ACE2 and NPC1 in SARS-CoV-2 entry. ACE2 mediates initial attachment and supports both TMPRSS2-dependent cell-surface fusion and endosomal uptake. By contrast, fusion within late endosomes requires NPC1, which interacts with the SARS-CoV-2 RBD through NPC1-C. These findings support a model in which NPC1 functions as a previously unrecognized intracellular receptor that mediates the endosomal entry pathway of SARS-CoV-2.

Results

SARS-CoV-2 (SARS2) entry shows reduced dependence on NPC1 in cells expressing TMPRSS2.

To define the role of NPC1 in viral infection, we generated NPC1-knockout (KO) variants of several cell lines, including Vero (African green monkey kidney), A549 (human alveolar basal epithelial adenocarcinoma) expressing ectopic human ACE2 (A549-A), Caco-2 (human colorectal adenocarcinoma), and Calu-3 (human lung adenocarcinoma). NPC1-KO was confirmed by Western blotting (WB) (Fig. 1A). These cells were infected with the SARS2 WA1 strain, and viral replication was quantified by plaque assay. Compared with their wild-type (WT) counterparts, NPC1-KO cells showed markedly reduced infection by 1,582-fold in Vero, 284-fold in A549-A, 38-fold in Caco-2, and 4.1-fold in Calu-3 cells. Thus, there is a cell-type-dependent requirement of NPC1 levels for SARS2 infection.

To determine whether this defect occurs at the viral entry step, we examined infected A549-A WT and NPC1-KO cells by transmission electron microscopy. In WT cells, viral particles were detected within endosome-like compartments, consistent with endosomal entry (Fig. 1 B, white arrow heads). However, NPC1-KO cells accumulated markedly higher numbers of virions within these compartments, indicating that viral entry proceeds to the endosomal stage but is blocked from progressing beyond it.

To test whether cholesterol transport contributes to NPC1-dependent infection, we knocked out NPC2—required for NPC1-mediated cholesterol export—in A549-A cells. Three independent NPC2-KO clones were established, with clone #3 showing complete loss of NPC2 (Fig. 1C). Infection with HIV-1–based pseudovirions (PVs) bearing SARS2 spike (S), EBOV-GP, or vesicular stomatitis virus glycoprotein (VSV-G) showed that all PVs infected both WT and NPC2-KO cells similarly, although SARS2-S PVs entered less efficiently than EBOV-GP or VSV-G PVs (Fig. 1D). Authentic SARS2 infection confirmed that NPC2 is not required for viral replication (Fig. 1E). Thus, NPC1 dependence of SARS2 infection is independent of cholesterol.

To investigate why SARS2 infection is far more dependent on NPC1 in Vero and A549-A than in Caco-2 and Calu-3, we assessed ACE2 and TMPRSS2 expression using flow cytometry (Fig. 1F) and WB (Fig. 1G). As controls, 293T and A549 cells were transduced with ACE2 (A) or ACE2 and TMPRSS2 (AT). Endogenous ACE2 expression was low in all parental lines, but high ACE2 expression was detected in transduced 293T-A, 293T-AT, A549-A, and A549-AT cells. Despite low ACE2 expression, Vero, Caco-2, and Calu-3 supported robust infection, indicating that high ACE2 levels are not essential for SARS2 infection. Endogenous TMPRSS2 expression, however, was high in Calu-3 and Caco-2, but low in A549 and Vero, and high TMPRSS2 expression was detected in transduced cells. Thus, cell lines that had high expression of endogenous TMPRSS2 (Caco-2, Calu-3) had the least reduction in viral replication upon NPC1-KO, suggesting that TMPRSS2 reduces the virus’s dependence on NPC1.

NPC1 is required for SARS2 endosomal entry but dispensable for cell-surface entry.

To precisely delineate NPC1’s role in these two entry pathways, we compared SARS2 infection in cells expressing ACE2 alone, which support productive endosomal entry, with cells expressing both ACE2 and TMPRSS2, which permit both endosomal and cell-surface entry. In addition to genetic NPC1 knockout, we treated cells with the NPC1 inhibitor Tubeimoside III (Tub-III). We also employed the cathepsin inhibitor E-64d, the TMPRSS2 inhibitor camostat, and the endosomal acidification inhibitor bafilomycin A1 (Baf-A1) to pharmacologically define NPC1-dependent entry mechanisms. E-64d inhibits SARS2 and EBOV endosomal entry, camostat blocks SARS2 surface entry, and Baf-A1 broadly inhibits viral endosomal entry. Both SARS2-S PV and authentic SARS2 were used for infection, with EBOV-GP and VSV-G PVs serving as controls.

Tub-III did not inhibit SARS2-S or EBOV-GP PV infection—or authentic SARS2 infection—in NPC1-KO cells (A549, HEK293T, and Vero), nor did it affect VSV-G PV infection (Fig. S1A–B), confirming its specificity for NPC1. SARS2-S PV infection was increased in A549-AT and Vero-AT cells compared with A549-A and Vero-A cells, consistent with enhanced surface entry mediated by ectopic TMPRSS2 expression (Fig. 2A). In contrast, TMPRSS2 expression had little effect on EBOV-GP or VSV-G PV infection.

Baf-A1 strongly inhibited EBOV-GP and VSV-G PV infection, whereas Tub-III selectively suppressed EBOV-GP PV infection regardless of TMPRSS2 expression (Fig. 2A). Notably, both inhibitors exhibited reduced inhibitory effects on SARS2-S PV infection in the presence of TMPRSS2. Authentic SARS2 infection recapitulated these results (Fig. 2B), indicating that Tub-III, like Baf-A1, specifically blocks the endosomal entry route.

Similarly, E-64d inhibited EBOV-GP PV infection independently of TMPRSS2 expression (Fig. S2A), consistent with its reliance on cathepsin activity. E-64d also strongly inhibited SARS2-S PV infection in A549-A, HEK293T-A, and Vero-A cells, as well as authentic SARS2 infection in A549-A and Vero-A cells, but not in their NPC1-KO counterparts or in cells expressing ectopic TMPRSS2, irrespective of NPC1 status (Fig. S2A–B). These results further support a role for NPC1 in the endosomal entry pathway. In contrast, camostat selectively inhibited SARS2-S PV and authentic SARS2 infection in TMPRSS2-expressing cells with similar potency in WT and NPC1-KO backgrounds, indicating that TMPRSS2-mediated surface entry is NPC1-independent.

Consistent with these pharmacological findings, NPC1-KO had no effect on VSV-G PV infection in A549 or Vero cells but markedly reduced EBOV-GP PV infection regardless of ACE2 or TMPRSS2 expression (Fig. 2C-D). NPC1-KO also impaired SARS2-S PV infection in A549-A and Vero-A cells, whereas this effect was substantially attenuated in A549-AT and Vero-AT cells. The same pattern was observed with authentic SARS2 infection (Fig. 2E). Notably, residual SARS2 infection persisted in NPC1-KO A549-A and Vero-A cells, particularly at 48 h post-infection (Fig. 2E), likely reflecting low-level surface entry mediated by endogenous TMPRSS2 or related proteases.

To directly assess whether NPC1 is required for surface fusion, we performed a split-NanoLuc (LgBiT/HiBiT) cell–cell fusion assay. Donor HEK293T cells expressing SARS2-S and LgBiT were co-cultured with recipient NPC1-KO HEK293T, 293T-A, or 293T-AT cells expressing HiBiT (Fig. 2F). TMPRSS2 expression markedly enhanced fusion, whereas NPC1-KO had little effect (Fig. 2G).

Together, these results demonstrate that NPC1 is dispensable for TMPRSS2-mediated surface entry but is specifically required for the endosomal route of SARS2 entry.

Omicron variant infection exhibits enhanced dependence on NPC1.

Consistent with prior reports that Omicron preferentially utilizes endosomal entry in cell culture ^11, 12, 13^, we observed that the Omicron spike induced markedly weaker cell–cell fusion than Delta or Gamma in an mCherry-based fusion assay (Fig. 3A). Moreover, whereas TMPRSS2 expression substantially enhanced Delta- and Gamma-mediated fusion, it had little effect on Omicron-mediated fusion. These findings were independently confirmed using the split-NanoLuc–based cell–cell fusion assay (Fig. 3B).

Pharmacological inhibition further supported this entry preference. Omicron PV and authentic virus infections were strongly inhibited by the cathepsin inhibitor E-64d but were largely insensitive to the TMPRSS2 inhibitor camostat. In contrast, Delta and Gamma infections were inhibited by both inhibitors (Fig. S3A–B). Together, these results indicate that Omicron is substantially less dependent on cell-surface entry than Delta and Gamma.

Consistent with this conclusion, Delta and Gamma PV infections increased by up to tenfold in ACE2/TMPRSS2-expressing cells compared with ACE2-only cells, whereas Omicron PV infection showed minimal enhancement (Fig. 3C). NPC1-KO significantly reduced Delta and Gamma PV infection in ACE2-only cells but had little effect in TMPRSS2-expressing cells. In striking contrast, Omicron PV infection was strongly suppressed by NPC1-KO even in TMPRSS2-expressing cells. Similarly, the NPC1 inhibitor Tub-III selectively inhibited Delta and Gamma PV infection in A549-A cells but not in A549-AT cells, while robustly suppressing Omicron PV infection irrespective of TMPRSS2 expression (Fig. 3D). Authentic Delta and Omicron virus infections yielded comparable results (Fig. 3E).

Collectively, these findings demonstrate that Omicron exhibits a markedly increased dependence on NPC1 relative to Delta and Gamma, consistent with its preferential reliance on the endosomal entry pathway in cell culture.

Detection of the SARS2-S-NPC1 complex in late endosomes and lysosomes.

To visualize interactions between SARS2-S and host entry factors, we employed a bimolecular fluorescence complementation (BiFC) assay. The bright yellow fluorescent protein Venus was split into non-fluorescent N- and C- terminal fragments, which were fused to either SARS2-S and ACE2 or SARS2-S and NPC1. EBOV-GP was included as a control.

Reconstituted Venus fluorescence was observed for ACE2–SARS2-S, NPC1–SARS2-S, and NPC1–EBOV-GP pairs, but not for ACE2–EBOV-GP, confirming the specificity of these interactions (Fig. 4). The NPC1–SARS2-S complex showed strong colocalization with the lysosomal marker LAMP1 and the late endosomal marker Rab7, as well as partial colocalization with the early endosomal marker Rab5. Similarly, the NPC1–EBOV-GP complex predominantly colocalized with LAMP1. In contrast, neither complex colocalized with the recycling endosome marker Rab11. Notably, the ACE2–SARS2-S complex did not colocalize with any endosomal or lysosomal markers examined.

Together, these results indicate that, unlike the ACE2–SARS2-S interaction, which occurs at the cell surface, the SARS2-S–NPC1 interaction predominantly takes place in late endosomes and lysosomes, closely resembling the intracellular localization of the NPC1–EBOV-GP complex.

Distinct surfaces of SARS2-RBD engage ACE2 at neutral pH and NPC1 under acidic conditions.

As early endosomes mature into late endosomes and lysosomes, progressive acidification occurs through compensatory counter-ion fluxes across the endosomal membrane ^14^. Under low-ionic-strength, acidic conditions (pH 6.0), biolayer interferometry demonstrated that NPC1 loop C (NPC1-C) binds SARS2-RBD with much higher affinity than at neutral pH (7.0) (Fig. 5A).

To validate this pH-dependent interaction under more physiological conditions, we purified SARS2-S PVs by ultracentrifugation. Following cleavage with the bacterial metalloprotease thermolysin—which mimics entry-related proteolysis—the PVs were incubated with purified ACE2 or NPC1 across a range of pH conditions. We found that cleaved SARS2-S pulled down ACE2 most efficiently at pH 7.5, with markedly reduced binding at pH 5.2. In contrast, NPC1 binding was the strongest at pH 5.2 and substantially decreased at pH 7.5 (Fig. 5B).

To delineate the binding interface, we performed competition assays with ACE2 and different classes of SARS2-neutralizing antibodies recognizing distinct regions on the RBD: class I antibody (ACE2-blocking, binding RBD in the “up” conformation) B1-182.1 and class II antibody (ACE2-blocking, binding RBD in both “up” and “down” conformations) A19-46.1 ^15^, as well as class IV antibodies (non-ACE2-blocking, binding RBD in the “up” conformation) A18-448.1 and F768-D11 (unpublished). The results showed that NPC1-C competes with both class I and class II antibodies, but not with class IV antibodies or ACE2 (Fig. 5C), indicating a putative NPC1-C binding site located between the epitopes of B1-182.1 and A19-46.1 and distinct from the ACE-binding site (Fig. 5D, left). This interaction pattern is consistent with the model predicted by AlphaFold3, in which NPC1-C engages a region on RBD overlapping with the epitopes of both class I and class II antibodies but away from the ACE2-binding site (Fig. 5D, right).

NPC1-C promotes SARS2-S–mediated membrane fusion under acidic conditions.

NPC1-like 1 (NPC1L1) shares ~ 40% sequence identity with NPC1 and has a similar membrane topology with three luminal loops (Fig. S4A-C). However, unlike NPC1—which localizes to late endosomes and lysosomes—NPC1L1 is found at the plasma membrane, where it facilitates cholesterol uptake ^16^. EPM is a synthetic coronavirus receptor comprising a 132–amino acid LCB1 binding module, an immunoglobulin domain, the MXRA8 transmembrane and cytosolic regions, and an endocytosis-prevention motif (EPM) (Fig. S5) ^17^

To test whether NPC1-C alone is sufficient to promote membrane fusion, we generated three chimeric receptors: NPC1L1-1C (NPC1L1 with its loop C replaced by NPC1-C) and EPM-1C and EPM-L1C (EPM with LCB1 replaced with NPC1-C or NPC1L1-C). Expression of all chimeras was confirmed by WB (Fig. S6).

We first used a split-NanoLuc (LgBiT/HiBiT) cell-free fusion assay involving virus-like particles (VLPs) and extracellular vesicles (EVs) ^18^. Fusion occurred efficiently between ACE2-LgBiT EVs and SARS2-HiBiT VLPs, and this activity was unchanged using EVs prepared from NPC1-KO cells (Fig. 6A, top), confirming that NPC1 is dispensable for cell surface fusion. Among the chimeras expressed on EVs, ACE2 exhibited the highest fusion activity, followed by EPM and EPM-1C, whereas EPM-L1C did not show any activity (Fig. 6A, bottom). These results indicate that NPC1-C, but not NPC1L1-C, can support SARS2-S–mediated fusion.

We next assessed how pH affects this fusion activity using a GFP/mCherry-based cell–cell fusion assay. Donor HEK293T cells expressing SARS2-S and GFP were co-cultured with recipient HEK293T cells expressing the chimeric proteins and mCherry. Consistently, at pH 7.0, ACE2 expression generated large double-positive syncytia, showing the strongest fusion activity (Fig. 6B-C, Fig. S7). EPM and EPM-1C exhibited intermediate activity, NPC1L1-1C supported modest fusion, whereas NPC1, NPC1L1, and EPM-L1C showed no detectable activity. Notably, reducing the pH to 5.0 had no effect on EPM-mediated fusion but markedly decreased ACE2-mediated fusion and, importantly, significantly enhanced EPM-1C–mediated fusion (Fig. 6D).

Collectively, these findings demonstrate that NPC1-C is a functional fusion module for SARS2-S under acidic conditions, supporting its essential role in viral entry through late endosomes/lysosomes.

A pH-dependent receptor switch model for SARS2.

SARS2-S trimers adopt two predominant prefusion conformations: an RBD-up ("open") conformation and an all-RBD-down ("closed") conformation, located within the S_1_ subunit ^19, 20, 21^. The "up" positioning of the RBD is required for interaction with the ACE2 receptor. The propensities of the open and closed conformations have been shown to vary with changes in pH, with the maximal open conformation (up to 68%) being detected at physiological pH 7.4 ^22^. This observation aligns with our previous finding showing that only the all-RBD-down conformation was detected at pH levels below 5 ^23^.

The progressive acidification that occurs during endosomal maturation may promote a shift to the RBD-down conformation, which conceals the ACE2-binding site while leaving the NPC1-C-accessible surface exposed. This supported a model in which progressive acidification accompanying endosomal maturation—from the nascent endocytic vesicle at the plasma membrane to the highly acidic late endosome—drives a receptor switch from ACE2 to NPC1. This framework provides a structural explanation for why NPC1–SARS2-S complexes are primarily detected in late endosomal and lysosomal compartments, accounting for their localization to these acidic organelles (Fig. 6E).

Discussion

Our study delineates a dynamic and context-dependent mechanism of SARS2 entry, highlighting a novel receptor switch from ACE2 to NPC1 during endocytic trafficking. While ACE2-mediated attachment and TMPRSS2-facilitated surface fusion are well-characterized pathways for viral entry, our findings reveal that NPC1 plays an essential role in endosomal fusion, especially under acidic conditions prevalent in late endosomes and lysosomes.

Structurally, the engagement of NPC1-C with the spike appears to compete with or complement neutralizing antibodies targeting class I and II epitopes, suggesting additional avenues for immune evasion and therapeutic targeting. Our chimeric receptor studies further confirm that NPC1-C alone can support fusion under acidic conditions, positioning NPC1 as a critical fusion facilitator within the endosomal compartment.

From a broader perspective, the identification of NPC1 as an intracellular receptor for SARS-CoV-2 expands the paradigm of coronavirus entry, placing lipid-trafficking machinery at the center of viral fusion processes. This insight opens new avenues for antiviral development—targeting NPC1 or its interaction with the spike could disrupt endosomal entry, especially in cell types that favor this pathway.

In conclusion, our work supports a model of SARS2 endosomal entry involving a pH-driven receptor switch from ACE2 to NPC1 (Fig. 6F). Such a mechanism underscores the virus’s adaptability in navigating host cell compartments and highlights the importance of intracellular factors in viral pathogenesis. Further structural and in vivo studies are warranted to exploit this receptor switch therapeutically and to understand its implications in different tissue contexts and disease outcomes.

Materials and Methods

Chemical reagents

Tubeimoside III (Cat. No. HY-N2542) and Bafilomycin A1 (Baf-A1; Cat. No. HY-100558) were purchased from MedChemExpress.

Anti-protease inhibitor cocktail (Cat. No. P8340) was obtained from Sigma-Aldrich.

Blasticidin S HCl (Cat. No. R21001) was from Thermo Fisher Scientific.

Puromycin (Cat. No. ant-pr-1), G418 (Cat. No. ant-gn-1), and Zeocin (Cat. No. ant-zn-1) were from InvivoGen.

Polyethyleneimine (Cat. No. 23966-1) was obtained from Polysciences.

E-64d (Cat. No. E1337-5MG) was from Fisher Scientific, and Camostat (Cat. No. 16018) was purchased from Cayman Chemical Company, Inc.

Glutaraldehyde 25% solution, EM Grade Distillation purified (Cat No. 16210) and Sodium cacodylate buffer 0.2M, pH 7.4 (Cat No. 11652) were purchased from Electron Microscopy Science.

Recombinant proteins

Recombinant SARS-CoV-2 RBD with a His-tag (Cat. No. 40592-V08H) and NPC1-C with His & Flag-tags (Cat. No. 16499-H32H) were purchased from SinoBiologicals. Recombinant soluble ACE2 with a FLAGtag (Cat. No. SAE0064) was purchased from Sigma.

Antibodies

Rabbit anti-SAR2-CoV-2 (2019-nCoV) Spike RBD antibodies (Cat. No: 40592-T62; 1:2000 for Western blotting) was purchased from Novus Biologicals.

Rabbit anti-human NPC1 antibody (Cat. No. ab134113; 1:5000 for Western blotting) was purchased from Abcam.

Mouse monoclonal anti-ACE2 IgG1 (Cat. No. 66699-1-Ig; 1:2000 for Western blotting, 1:100 for flow cytometry), rabbit polyclonal anti-NPC2 (Cat. No. 19888-1-AP; 1:2000), and rabbit polyclonal anti-TMPRSS2 (Cat. No. 14437-1-AP; 1:1000 for Western blotting) were obtained from Proteintech.

Rabbit polyclonal anti-TMPRSS2 (Cat. No. PA5-14264; 1:50 for flow cytometry) was from Invitrogen.

Goat anti-human ACE2 affinity-purified IgG (Cat. No. AF933; 1:1000 for Western blotting) was purchased from R&D Systems.

Rabbit polyclonal anti–SARS-CoV-2 spike protein (Cat. No. TA890227; 1:2000 for Western blotting) was from OriGene.

SARS-CoV-2 spike guinea pig antibody (Cat. No. NR-10361; 1:15,000 for plaque assay) was obtained from BEI Resources.

Horseradish peroxidase (HRP)–conjugated goat anti–guinea pig IgG (Cat. No. A-7289; 1:5000 for plaque assay) and HRP-conjugated anti-HA (Cat. No. H6533; 1:5000 for Western blotting) were purchased from Sigma-Aldrich.

Mouse monoclonal anti-LgBiT (Cat. No. N7100) was from Promega.

HRP-conjugated goat anti-mouse IgG (Cat. No. 115-035-003; 1:5000 for Western blotting) and HRP-conjugated goat anti-rabbit IgG (Cat. No. 111-035-003; 1:5000 for Western blotting) were purchased from Jackson ImmunoResearch.

Plasmids and molecular cloning

The following plasmids were reported previously ^10^: pNL4.3-ΔEnv-Luc, pcDNA3.1-VSV-G, pcDNA3.1-EBOV-GPΔMLD, pcDNA3.1-NPC1-3×FLAG, pCAGGS-SARS2-S-ΔC19-FLAG, pCAGGS-SARS2-S-Delta, pCAGGS-SARS2-S-Omicron, pCAGGS-SARS2-S-Gamma, and pLenti-BSD-hTMPRSS2-FLAG.

The lentiviral packaging plasmids pCMV-VSV-G (Addgene #8454; gift from Bob Weinberg) and p8.91 (Addgene #187441; gift from Simon Davis) were used for pseudovirus production. The CRISPR/Cas9 lentiviral expression vector LentiCRISPR-v2 (Addgene #52961) was from Feng Zhang.

Fluorescent reporters EGFP (Addgene #176015) and mCherry (Addgene #176016) were gifts from Rob Parton.

pDONR221-NPC1L1 (Addgene #132303) was from Giulio Superti-Furga.

Endosomal markers LAMP1-RFP (Addgene #1817), DsRed-Rab7 (Addgene #12661), DsRed-Rab11 (Addgene #12679), and mRFP-Rab5 (Addgene #14437) were from Walther Mothes and Richard Pagano.

LgBiT-mCherry (Addgene #199715) was from Alice Ting.

Constructs pcDNA3.1-hACE2-LgBiT (#215837), pcDNA3.1-SARS-CoV-2-S (Wuhan) (#215812), pcDNA3.1-SARS-CoV-2-M (#215813), pcDNA3.1-SARS-CoV-2-E (#215814), and pcDNA3.1-SARS-CoV-2-HiBiT-N (#215815) were from Thomas Gallagher via Addgene.

HA-Rab7A was from Feng-Qian Li and Ken-Ichi Takemaru via Addgene (#131417).

Plasmids encoding proteins of interest were generated using the Hieff Clone^™^ Universal II One-Step Cloning Kit (Yeasen, Cat. No. 10923-A) according to the manufacturer’s instructions.

The EPM sequence was described previously ^17^ and synthesized by Twist Bioscience. The EPM-1C construct was designed by replacing the LCB1 domain of EPM with the luminal loop C of NPC1 (NPC1-C) and synthesized by Twist Bioscience. NPC1 was amplified by PCR from pcDNA3.1-NPC1-3×FLAG, and NPC1L1 was amplified from pDONR221-NPC1L1. NPC1L1–NPC1-C was generated by substituting loop C of NPC1L1 (NPC1L1-C) with NPC1-C. Similarly, EPM-L1C was constructed by replacing NPC1-C in EPM–NPC1-C with NPC1L1-C.

pcDNA3.1-hACE2-LgBiT was used as the vector backbone. The corresponding constructs—NPC1-LgBiT, NPC1L1-LgBiT, NPC1L1–1C-LgBiT, EPM-LgBiT, EPM-1C-LgBiT, and EPM-L1C-LgBiT—were generated by replacing the hACE2 insert in pcDNA3.1-hACE2-LgBiT with the indicated genes.

The BiFC expression plasmids pcDNA3.1-mSer5-VN-HA and pcDNA3.1-mSer5-FLAG-VC were reported previously ^24^. NPC1, ACE2, and SARS2-S were amplified by PCR from pcDNA3.1-NPC1–3×FLAG, pcDNA3.1-hACE2-LgBiT, and pCAGGS-SARS2-S-ΔC19-FLAG, respectively. Using these vectors as backbones, pcDNA3.1-ACE2-VN-HA, pcDNA3.1-NPC1-VN-HA, and pcDNA3.1-SARS2-S-VN-HA were generated by replacing the mSer5 sequence in pcDNA3.1-mSer5-VN-HA with ACE2, NPC1, or SARS2-S, respectively. Likewise, pcDNA3.1-ACE2-FLAG-VC and pcDNA3.1-NPC1-FLAG-VC were constructed by replacing mSer5 in pcDNA3.1-mSer5-FLAG-VC with ACE2 or NPC1.

Ebola GP was amplified by PCR from pcDNA3.1-EBOV-GPΔMLD, and VC was amplified from pcDNA3.1-mSer5-FLAG-VC. The HA-Rab7A vector was digested with HindIII and XbaI and used as the backbone to generate pCS2-EBOV-GP-VC by fusing Ebola GP and VC.

Detailed information on vector construction is available upon request. All primers used for cloning are listed in Supplementary Table 1.

Cells

HEK293T (Cat. No. CRL-3216), Caco-2 (Cat. No. HTB-37), Calu-3 (Cat. No. HTB-55), and Vero-E6 (Cat. No. CRL-1586) cell lines were obtained from the American Type Culture Collection (ATCC). The following cell lines were obtained from BEI Resources: 293T-ACE2 (Cat. No. NR-52511), 293T-ACE2-TMPRSS2 (Cat. No. NR-55293), A549 (Cat. No. NR-52268), A549-ACE2 (Cat. No. NR-53821), A549-ACE2-TMPRSS2 (Cat. No. NR-59471), Vero-ACE2 (Cat. No. NR-53726), and Vero-ACE2-TMPRSS2 (Cat. No. NR-54970).

All cell lines were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin and cultured at 37°C in a humidified atmosphere containing 5% CO_2_.

Generation of NPC1- and NPC2-knockout (KO) cell lines

Small guide RNA (sgRNA) oligonucleotides targeting the NPC1 and NPC2 genes in human and Vero-E6 cells (listed in Supplementary Table S2) were designed ^25^ and cloned into the LentiCRISPR-v2 vector following BsmBI digestion. Lentiviral particles were produced in HEK293T cells by transfecting LentiCRISPR-v2, pCMV-Δ8.9, and pVSV-G plasmids at a ratio of 4:3:2 using polyethyleneimine (PEI).

Target cells were transduced with lentiviral supernatants, and after 48 hours, selection was initiated using appropriate antibiotics: puromycin (5 μg/mL) for 293T, 293T-ACE2, 293T-ACE2-TMPRSS2, A549, A549-ACE2, and Vero cells; puromycin (2 μg/mL) for Calu-3 cells; puromycin (5 μg/mL) for A549-ACE2 NPC2-KO cells; blasticidin (2 μg/mL) and G418 (1 mg/mL) for Vero-ACE2 NPC1-KO cells; zeocin (200 μg/mL) for Vero-ACE2-TMPRSS2 and A549-ACE2-TMPRSS2 NPC1-KO cells.

Single-cell clones were isolated in 96-well plates using a Beckman Coulter MoFlo Astrios EQ cell sorter. The Calu-3 NPC1-KO cell population was used directly after puromycin selection without clonal isolation.

Successful knockout of NPC1 and NPC2 was confirmed by Western blotting.

Pseudovirion production

HEK293T cells were seeded in 10-cm dishes and transfected with 10 μg of HIV-1 proviral vector (pNL4.3-ΔEnv-Luc) and 2.5 μg of viral spike or glycoprotein expression plasmid (EBOV-GP, SARS2-S, or VSV-G) using polyethyleneimine (PEI). Supernatants were harvested 48 h post-transfection and clarified twice by centrifugation at 3,000 × g for 10 min to remove cellular debris, followed by filtration through a 0.22 μm membrane. The clarified pseudovirions were aliquoted and stored at − 80°C until use in infection experiments. Infectivity was calculated after being normalized by p24^Gag^ ELISA ^26^.

Flow cytometry

Approximately 2 × 10^6^ cells from each line were collected and washed twice with phosphate-buffered saline (PBS). Cells were incubated with primary antibodies against ACE2 (Proteintech, Cat. No. 66699-1-Ig, 1:100) and TMPRSS2 (Invitrogen, Cat. No. PA5-14264, 1:50) for 30 min at 4°C in the dark. After washing with PBS containing 1% bovine serum albumin (BSA), cells were stained with PE-conjugated donkey anti-rabbit IgG (BioLegend, Cat. No. 405308, 1:200) and APC-conjugated goat anti-mouse IgG (BioLegend, Cat. No. 405308, 1:100) under identical conditions. Nuclei were counterstained with DAPI (1 μg/mL) for 5 min. After a final PBS wash, cells were resuspended in PBS and analyzed on a CytoFLEX S flow cytometer (Beckman Coulter). Background staining was assessed using mouse IgG1 (Proteintech, Cat. No. 66360-1-Ig, 1:500) and rabbit IgG (Proteintech, Cat. No. 30000-0-AP, 1:1000). Data were analyzed with FlowJo v10.10.

Western blotting (WB)

Cells were lysed in ice-cold RIPA buffer (25 mM Tris-HCl, pH 7.4; 150 mM NaCl; 0.5% sodium deoxycholate; 0.1% SDS; 1% NP-40; Sigma-Aldrich, R0278) supplemented with a protease inhibitor cocktail (Sigma-Aldrich, P8340). Approximately 0.1 mL RIPA buffer was used per 2 × 10^6^ cells. Lysates were clarified by centrifugation at 12,000 × g for 10 min at 4°C, and supernatants were boiled in SDS-PAGE loading buffer (Solarbio P1015). Proteins were resolved by SDS-PAGE, transferred to PVDF membranes, and blocked with 5% non-fat milk in TBST (20 mM Tris, pH 7.4; 150 mM NaCl; 0.1% Tween-20) for 1 h at room temperature. Membranes were incubated with primary antibodies overnight at 4°C, washed four times (20 min each) in TBST, and then incubated with HRP-conjugated secondary antibodies (1:5000) for 1 h. Signals were detected using Immobilon Classico Western HRP Substrate (Millipore, WBLUC0500) ^27^ and imaged on an iBright 1500 Imager (Invitrogen).

Authentic SARS-CoV-2 infection

Authentic SARS-CoV-2 strains included USA-WA1/2020 (WA1) ^28^, hCoV-19/USA/MD-HP05285/2021 (Delta; B.1.617.2) (BEI Resources, NR-55671), and hCoV-19/USA/COR-22-063113/2022 (Omicron; BA.5) (BEI Resources, NR-58616).

All infections were performed in a biosafety level 3 (BSL-3) facility at the University of Illinois at Chicago. Cells were infected at a multiplicity of infection (MOI = 0.01) for 1 h, washed three times with cold PBS containing 2% FBS, and cultured for 24 h. Viral titers were determined by plaque assay on Vero-E6 cells.

Plaque assay

Vero cells were seeded at 2 × 10^4^ cells/well in 96-well plates. The next day, virus samples were serially diluted 10-fold in DMEM + 2% FBS and 100 μL of each dilution was added to cells for 1 h at 37°C. After removing inocula, cells were overlaid with 1% methylcellulose and incubated 24 h at 37°C. Cells were fixed with 4% paraformaldehyde for 30 min, washed, and immunostained using SARS2-spike guinea pig antibody (BEI NR-10361; 1:15,000) followed by anti-guinea pig HRP (Sigma A-7289; 1:5000). Spots were visualized using TrueBlue Substrate (Seracare 5510-0030). Plates were washed, dried, and imaged using a Biospot Plate Reader (ImmunoSpot 7.0.15.2).

Cell–cell Fusion Assays

For ACE2–SARS2-S–mediated fusion, HEK293T cells were transfected with LgBiT-mCherry + SARS2-S or Vpr-HiBiT constructs. At 24 h post-transfection, SARS2-S-expressing donor cells were mixed 1:1 with HEK293T, 293T-ACE2, or 293T-ACE2-TMPRSS2 cells. After 48 h co-culture, fusion was assessed by mCherry imaging (KEYENCE BZ-X810) or quantified using the Nano-Glo Live Assay Kit, calculating fold increases of NanoLuc signal over ACE2-negative controls.

For receptor-specific fusion, HEK293T cells expressing EGFP + SARS2-S (recipients) were co-cultured 1:1 with HEK293T cells expressing mCherry + receptor constructs (NPC1, NPC1L1, NPC1L1-NPC1-C, ACE2, EPM, EPM-NPC1-C, EPM-NPC1L1-C). After cell adhesion (6 h), the medium was replaced with DMEM pH 5.0 (adjusted with HCl) for low-pH treatment. At 24 h, double-positive cells were analyzed by flow cytometry; at 48 h, fused cells were imaged microscopically.

Virus-like particle (VLP) production

HiBiT-N–tagged VLPs were produced by co-transfecting HEK293T cells with plasmids encoding S, E, M, and HiBiT-N using PEI as reported ^18^. “No S” VLPs were generated using an empty vector in place of S. After 16 h, media were replaced with fresh DMEM + 10% FBS. VLPs were collected from FBS-free DMEM 24–48 h post-transfection and clarified by sequential centrifugation (300 × g, 10 min; 3,000 × g, 10 min). Supernatants were concentrated 100-fold by ultrafiltration (Millipore UFC910096) and purified by sizeexclusion chromatography (SEC) (Izon AFC-V2) using PBS (pH 7.4). Peak VLP fractions were identified by NanoLuc complementation with LgBiT-containing lysates and normalized by luminescence. Samples were stored at − 80°C.

Cell-free fusion assay

Extracellular vesicles (EVs) carrying various proteins were obtained by transfecting HEK293T cells with pcDNA3.1-hACE2-LgBiT, pcDNA3.1-hACE2-LgBiT, NPC1L1-LgBiT, NPC1L1-1C-LgBiT, EPM-LgBiT, EPM-1C-LgBiT, EPM-L1C-LgBiT, or hNPC1-LgBiT as reported ^18^. Media were collected 48 h post-transfection, clarified, concentrated 100-fold, and purified by SEC (Izon AFC-V2). Peak EV fractions were stored at 4°C.

For fusion assays, HiBiT-N VLPs and ACE2-LgBiT EVs were combined with NanoLuc substrate (Promega N2420), 10 μM DrkBiT (NovoPro 319394), and trypsin (1 ng/μL) in 96-well plates. Luminescence was recorded over time using a Tecan Infinite 200 Pro luminometer at 37°C. Fusion was quantified as the fold increase in signal relative to spikeless (No S) VLP controls.

Bimolecular fluorescence complementation (BiFC)

HeLa cells were seeded on glass coverslips in 12-well plates and transfected with PEI. The following plasmid pairs were used (0.8 μg each):

pCS2-EBOV-GP-VC + pcDNA3.1-ACE2-VN-HA (EBOV-GP–ACE2)pCS2-EBOV-GP-VC + pcDNA3.1-NPC1-VN-HA (EBOV-GP–NPC1)pcDNA3.1-ACE2-FLAG-VC + pcDNA3.1-SARS2-S-VN-HA (SARS2-S–ACE2)pcDNA3.1-NPC1-FLAG-VC + pcDNA3.1-SARS2-S-VN-HA (SARS2-S–NPC1)

Organelle markers were co-transfected as follows: LAMP1-RFP (lysosomes), DsRed-Rab7 (late endosomes), DsRed-Rab11 (recycling endosomes), and mRFP-Rab5 (early endosomes) (0.4 μg each).

At 72 h post-transfection, cells were fixed in 4% paraformaldehyde for 15 min, counterstained with DAPI, and mounted with ProLong Gold Antifade Mountant (Invitrogen P10114). Reconstituted Venus fluorescence was imaged on a Zeiss LSM 710 confocal microscope, and analyzed using ZEN 3.11 software.

Transmission electron microscope (TEM)

Confluent A549-A WT and NPC1-KO cells were grown in 10-cm dishes and infected with SARS-CoV-2 Omicron at a multiplicity of infection (MOI) of 40 for 12 h. After infection, the culture medium was removed, and 2.5% buffered glutaraldehyde (2.5% glutaraldehyde in 0.1 M sodium phosphate buffer) was gently added along the side of each dish. Cells were fixed for 1 h at room temperature and then scraped and transferred into 1.5-mL Eppendorf tubes. Samples were pelleted at 300 × g for 10 min.

Fixed pellets were submitted to the Electron Microscope Center–West at the University of Illinois Chicago for processing and grid preparation. Briefly, samples were rinsed with double-distilled water and dehydrated through a graded ethanol series (30%, 50%, 70%, 80%, and 90%). Cell pellets were then infiltrated and embedded in fresh 100% LR White resin using BEEM capsules. Ultrathin sections (80 nm) were cut using a Leica UCT ultramicrotome and mounted on 400-mesh high-transmission grids.

For immunogold labeling, sections were incubated with rabbit anti-ALIX antibody followed by goat anti-rabbit IgG conjugated to 10-nm colloidal gold particles, and subsequently counterstained with uranyl acetate. Images were acquired using a JEOL JEM-1400 Flash transmission electron microscope equipped with a side-mount NanoSprint 1200S-B camera and AMT imaging software (v7.01).

Detection of SARS2-S interaction with ACE2 and NPC1 under different pH conditions

Purified SARS-CoV-2 pseudovirions (SARS2-HA) were cleaved with thermolysin (Sigma P1512) in 3 M sodium acetate buffer (pH 5.2; Beyotime ST351) for 1 h at 37°C, followed by inhibition with phosphoramidon (Beyotime SG2024). Cleaved proteins were incubated overnight with anti-HA affinity gel at 4°C. The bound spike protein was washed extensively in acetate buffers of varying pH to test pH stability, then incubated overnight with lysates containing NPC1-FLAG or ACE2-His. After five PBS washes, bound proteins were analyzed by WB using anti-HA-HRP (Sigma H6533), anti-NPC1 (Abcam ab134113), and anti-ACE2 (R&D Systems AF933).

Octet assessment

The binding of NPC1-C to the SARS2-RBD was evaluated using a FortéBio Octet Red384 instrument. All assays were performed at 30°C with agitation set to 1,000 rpm in 20 mM phosphate buffer (pH 6.0 or 7.0). Streptavidin (SA) biosensor tips were equilibrated in buffer for 60 s before loading biotinylated RBD. The biosensors were subsequently dipped into solutions containing 100 μg/mL of antibodies B1-182.1, F768-D11, A19-46.1, A18-448.1, or receptor ACE2, followed by immersion in 500 μg/mL of NPC1-C. Data acquisition and analysis were performed using Octet software, version 10.0.

Expression of monoclonal antibodies

For antibody expression, equal amounts of heavy and light chain plasmid DNA were transfected into Expi293 cells (Life Technology) by using Expi293 transfection reagent (Life Technology). The transfected cells were cultured in shaker incubator at 120 rpm, 37°C, 9% CO_2_ for 4 ~ 5 days. Culture supernatants were harvested and filtered, mAbs were purified over Protein A (GE Health Science) columns. Each antibody was eluted with IgG elution buffer (Pierce) and immediately neutralized with one tenth volume of 1M Tris-HCL pH 8.0. The antibodies were then buffer exchanged as least twice in PBS by dialysis.

AlphaFold3 modelling of the NPC1-C and SARS2 RBD interaction

Models of the NPC1-C in complex with SARS2 RBD were performed on the AlphaFold3 server (https://alphafoldserver.com/). The accuracy of the predicted relative positions of the subunits within the complex interface is measured by the predicted template modeling (ipTM). ipTM values higher than 0.8 represent confident high-quality predictions. In this study, predication run had a lower ipTM value lower than 0.8. However, the predicted binding mode shown in Fig. 5D, located between epitopes of class I and II RBD antibodies, is supported by antibody competition data shown in Fig. 5C.

Graphic Preparation

All fiures were prepared using Adobe Illustrator 2021. Schematic models were created using BioRender (https://biorender.com) and PyMOL (https://www.pymol.org).

Supplementary Material

This is a list of supplementary files associated with this preprint. Click to download.

• Supplementals.pdf