SARS-CoV-2 Omicron spike H655Y mutation is responsible for enhancement of the endosomal entry pathway and reduction of cell surface entry pathways

The SARS-CoV-2 Omicron variant reportedly displays decreased usage of the cell surface entry pathway mediated by the host transmembrane protease, serine 2 (TMPRSS2) and increased usage of the endosomal entry pathway mediated by cathepsin B/L. These differences result in different cell tropisms and low fusogenicity from other SARS-CoV-2 variants. Recent studies have revealed that host metalloproteases are also involved in cell surface entry and fusogenic activity of SARS-CoV-2, independent of TMPRSS2. However, the involvement of metalloproteinase-mediated cell entry and fusogenicity in Omicron infections has not been investigated. Here, we report that Omicron infection is less sensitive to the metalloproteinase inhibitor marimastat, like the TMPRSS2 inhibitor nafamostat, and is more sensitive to the cathepsin B/L inhibitor E-64d than infections with wild-type SARS-CoV-2 and other variants. The findings indicate that Omicron preferentially utilizes the endosomal pathway rather than cell surface pathways for entry. Moreover, the Omicron variant also displays poor syncytia formation mediated by metalloproteinases, even when the S cleavage status mediated by fusion-like proteases is unchanged. Intriguingly, the pseudovirus assay showed that a single mutation, H655Y, of the Omicron spike (S) is responsible for the preferential entry pathway usage without affecting the S cleavage status. These findings suggest that the Omicron variant has altered entry properties and fusogenicity, probably through the H655Y mutation in its S protein, leading to modulations of tissue and cell tropism, and reduced pathogenicity. Author summary Recent studies have suggested that the SARS-CoV-2 Omicron variant displays altered cell tropism and fusogenicity, in addition to immune escape. However, comprehensive analyses of the usage of viral entry pathways in Omicron variant have not been performed. Here, we used protease inhibitors to block each viral entry pathway mediated by the three host proteases (cathepsin B/L, TMPRSS2, and metalloproteinases) in various cell types. The results clearly indicated that Omicron exhibits enhanced cathepsin B/L-dependent endosome entry and reduced metalloproteinase-dependent and TMPRSS2-dependent cell surface entry. Furthermore, the H655Y mutation of Omicron S determines the relative usage of the three entry pathways without affecting S cleavage by the host furin-like proteases. Comparative data among SARS-CoV-2 variants, including Omicron, may clarify the biological and pathological phenotypes of Omicron but increase the understanding of disease progression in infections with other SARS-CoV-2 variants.


Introduction
The Omicron variant of severe acute respiratory syndrome-associated coronavirus-2 (SARS-CoV-2), the causative virus of coronavirus disease 2019 (COVID- 19), has spread globally since it was recognized as a variant of concern (VOC) by the World Health Organization on November 24, 2021. Despite its high infectivity, the pathogenicity of Omicron seems to be lower than that of other SARS-CoV-2 variants [1][2][3][4]. The Omicron variant has a large number of mutations in the viral genome, especially in the spike (S) region, which appears to contribute to the rapidly increasing number of cases worldwide by evading host immune responses in vaccinated or previously exposed individuals. Several research groups have reported that Omicron escapes from the neutralizing activities of therapeutic monoclonal antibodies and vaccine-elicited antibodies [5][6][7][8][9]. In addition to immune response escape, Omicron appears to exhibit different cell tropism from other SARS-CoV-2 variants [10][11][12][13], which may also contribute to the rapid spread of Omicron infection.
The S protein of SARS-CoV-2 is essential for viral entry into target cells. Prior to viral entry, the S protein is cleaved into S1 and S2 by host furin-like proteases before the release of viral particles from infected cells [14][15][16]. The first step is the binding of the S protein to the cell surface receptor angiotensin-converting enzyme 2 (ACE2) through the receptor binding domain (RBD) in the S1 domain [17,18]. In the second step, the S2 domain is further cleaved into S2' by either transmembrane serine protease 2 (TMPRSS2) localized in the plasma membrane or 5 cathepsin B/L in endosomal vesicles. This step is called priming [19,20]. S2' priming causes membrane fusion between the viral particle and the plasma or endosomal membrane through interaction between the fusion peptide region and the cellular membrane, resulting in the release of viral RNA into the cytoplasm and its replication. We and others recently showed that host metalloproteinases are also involved in cell surface entry, independent of transmembrane protease, serine 2 (TMPRSS2) [21][22][23]. These results suggest that metalloproteinases act as the third host proteases for SARS-CoV-2 entry. Additionally, the S protein localized in the cell membrane is primed by TMPRSS2 or metalloproteinases after binding to ACE2 on the surface of neighboring cells, resulting in cell-cell fusion to generate syncytia. Syncytia formation is attributed to tissue damage [24][25][26].
A difference in the use of entry pathways has been observed among SARS-CoV-2 variants. The Delta variant, which was the predominant SARS-CoV-2 strain before the spread of Omicron, displays increased usage of the TMPRSS2-mediated cell surface entry pathway compared to wild-type (WT) SARS-CoV-2 [26]. In addition, the Delta variant shows enhanced syncytia formation, which may be related to its more severe pathogenicity compared to other variants [26,27]. In contrast, the Omicron variant shows decreased usage of TMPRSS2dependent cell surface entry [2,10,12,13]. Furthermore, Omicron has low fusogenicity mediated by TMPRSS2, forming smaller syncytia than WT and Delta viruses [2,10,12,13]. The reduction 6 in TMPRSS2-mediated fusogenic activity may partially explain the low pathogenicity of the Omicron variant. However, the involvement of metalloproteinase-mediated viral entry and syncytia formation in Omicron infections has not been fully elucidated. In addition, the function of Omicron S mutations on entry pathway usage and fusogenicity remains unknown.
In this study, we investigated the relative usage of the three entry pathwaysmetalloproteinase-, TMPRSS2-, and cathepsin B/L-dependent pathway-by the Omicron variant.
Our comprehensive analyses showed that Omicron preferentially utilized the cathepsin B/Ldependent endosomal entry pathway rather than the metalloproteinase-and TMPRSS2-dependent cell surface entry pathways in various cell types. Furthermore, syncytia formation mediated by metalloproteinases was significantly reduced, similar to that by TMPRSS2. Finally, H665Y of the Omicron S protein was identified as the mutation responsible for determining the preferential usage of the endosomal pathway without affecting S1/S2 cleavage status.

Omicron variant preferentially uses the cathepsin B/L-dependent endosomal pathway over metalloproteinase-or TMPRSS2-dependent cell surface pathway
To evaluate the usage of the three different entry pathways by the SARS-CoV-2 variants, we used three protease inhibitors: nafamostat, marimastat, and E-64d. Nafamostat is a serine protease 7 inhibitor that specifically blocks TMPRSS2-dependent (TMPRSS2 pathway) cell surface entry [28,29]. Marimastat is a broad metalloproteinase inhibitor that blocks metalloproteinasedependent (metalloproteinase pathway) cell surface entry [21]. E-64d is a cathepsin B/L inhibitor that suppresses the cathepsin B/L-dependent endosomal entry pathway (endosomal pathway) [19]. Omicron infection was much more dependent on the endosomal pathway than the WT and Delta infections (Fig. 1B). These data show that the metalloproteinase pathway is almost completely lost in Omicron infection, and that this variant significantly relies on the endosomal pathway.
Next, we analyzed entry of Omicron into cells via the TMPRSS2 pathway. We tested VeroE6 cells ectopically expressing TMPRSS2 (VeroE6-TMPRSS2 (ATCC)), which predominantly depends on the TMPRSS2 pathway for WT infection (SFig. 1A). Although WT and Delta predominantly used the TMPRSS2 pathway in these cells, Omicron entry was significantly dependent on the 9 endosomal pathway (Fig. 1B). We further evaluated the effects of protease inhibitors on the viral infection of HEC50B cells ectopically expressing TMPRSS2 (HEC50B-TMPRSS2), which have all three entry pathways for WT infection (Fig. 1B). In Omicron infection, the contribution of the endosomal pathway was enhanced, while the TMPRSS2 and metalloproteinase pathways were less preferentially utilized compared to WT and Delta (Fig. 1B). These findings suggest that 2). In addition, increased usage of the endosomal pathway by the Omicron S pseudovirus was observed in both VeroE6-TMPRSS2 (ATCC) cells ( Fig 1C) and Caco-2 (colon cells) harboring the TMPRSS2 and endosomal pathways (SFig. 2). Taken together, these results suggest that Omicron S preferentially uses the cathepsin B/L-dependent endosomal entry pathway, rather than metalloproteinase-and TMPRSS2-dependent cell surface entry pathways. Interestingly, there was no difference in the dose-dependency of the inhibitory effects of nafamostat between Omicron and the other variants in Calu-3 cells (SFig.1 C and SFig. 2). In OVISE cells, WT and Omicron infections showed similar sensitivities to E-64d (SFig. 2). These data suggest that Omicron uses the same entry pathway as WT when infecting cells that have only one entry pathway.
Cleavage of SARS-CoV-2 S protein by host furin-like proteases affects the entry pathways for SARS-CoV-2 infection [2,12,13,26,30]. Therefore, we investigated the cleavage status of the S proteins in the variants. Western blot analysis of the virions using an anti-S2 domain antibody showed that the S protein of the Alpha variant was cleaved to the same degree as the WT S, while the cleavage efficiency of the Delta S was higher than that of the WT and Alpha S (Fig. 1D). Interestingly, there was no significant difference in the cleavage efficiency between Omicron S and WT S (Fig. 1D). These results suggest that Omicron S can change the relative usage of entry pathways, even when Omicron S is cleaved as efficiently as WT S.

Omicron variant displays poor syncytium formation mediated by metalloproteinases
without affecting the S1/S2 cleavage of its S protein Reduced syncytia formation through the TMPRSS2 pathway has been reported in Omicron, suggesting a correlation with lower pathogenicity [2,10,12,13]. We previously reported that metalloproteinases can also mediate syncytia formation by WT virus in HEC50B cells [21]. Thus, we examined the ability of Omicron to form syncytia via metalloproteinases in these cells. After 24 h of infection with WT, Omicron, or Delta, the cells were stained with anti-SARS-CoV-2 N 11 protein to observe the formation of syncytia. WT infection formed numerous syncytia both at a low MOI of 0.1 and a high MOI of 1 ( Fig. 2A and SFig. 3). Consistent with our previous report [21], metalloproteinase-mediated syncytia formation was observed in WT infection at both MOI values (SFig. 3). However, Omicron infection induced much smaller and fewer syncytia than WT infection, even when the cells were infected at the high MOI of 1 ( Fig. 2A). Although Delta infection also induced syncytia formation in a metalloproteinase-dependent manner in HEC50B cells (SFig. 3), the syncytia that formed where larger than those in WT infection ( Fig. 2A), as enhanced syncytia formation in TMPRSS2 expressing cells reported previously [26,27]. As Omicron replication was less than that of WT virus in HEC50B cells (Fig. 1A), it is possible that poor syncytia formation by Omicron was due to its reduced replication in HEC50B cells.
Therefore, we measured the intracellular viral RNA levels when syncytia formation was observed.
Omicron infection at an MOI of 1 resulted in much poorer syncytia formation than WT infection at an MOI of 0.1, even when the Omicron infection produced more viral RNA in the cells than the WT infection (Fig. 2B). In contrast, Delta infection produced more syncytia than WT and Omicron infections, regardless of the amount of viral RNA (Fig. 2B). These results indicated that the Omicron variant has a poor ability to form syncytia by TMPRSS2 and metalloproteinases, and that the Delta variant induces enhanced syncytia formation by metalloproteinases.
Several studies reported that S1/S2 cleavage of the SARS-CoV-2 S protein expressed in 12 infected cells affects syncytium formation [2,12,13]. However, in the present study, there was no significant difference in the S1/S2 cleavage status between the WT, Alpha, and Omicron variants in HEC50B cells, although Delta S was cleaved more than the S proteins of the other variants (Fig. 2C), as previously reported [26]. These results indicate that Omicron forms syncytia poorly, even when S cleavage status is similar to that of WT.

H655Y mutation of Omicron spike protein determines preferential usage of endosomal pathway without affecting S1/S2 cleavage status
To elucidate the molecular mechanism underlying the preferential usage of the endosomal pathway by Omicron, we attempted to identify the responsible mutation sites in the S protein. The Omicron BA.1 variant (TY38-873) has 32 amino acid mutations in its S protein compared to the WT, with 24 mutations in the S1 region, including the RBD, two mutations in the S1/S2 boundary region including the furin cleavage site, and six mutations in the S2 region (Fig. 3A). We generated various pseudoviruses with chimeric S proteins that consisted of the S1, S1/S2 boundary, and S2 regions derived from either WT or Omicron (Fig. 3B). In HEC50B cells, pseudoviruses bearing the S proteins consisting of the WT S1 region (WOO, WOW, and WWO) infected the cells mainly through the metalloproteinase pathway (Fig. 3C), similar to the WT pseudovirus. In contrast, the pseudoviruses bearing the S proteins consisting of the Omicron S1 13 region (OWO, OOW, and OWW) infected the cells mainly through the endosomal pathway (Fig.   3C). Furthermore, in VeroE6-TMPRSS2 (ATCC) cells, the WOO, WOW, and WWO pseudoviruses predominantly used the TMPRSS2 pathway, similar to the WT pseudovirus, whereas the OWO, OOW, and OWW pseudoviruses showed the same pattern of entry pathway usage as the Omicron S pseudovirus (Fig. 3C). These results clearly demonstrated that the S1 region of the Omicron S protein is responsible for preferential usage of the endosomal pathway.
To identify the amino acid mutations in the S protein responsible for the preferential usage of the endosomal pathway by Omicron, we generated 24 pseudoviruses bearing Omicron S proteins, with each mutated amino acid residue in the Omicron S1 region substituted for the corresponding amino acid in the WT S1 region. Among them, the Omicron S protein with an amino acid substitution at position 373 (373S) failed to produce infectious virus particles. VeroE6-TMPRSS2 (ATCC) cells (Fig. 3D). These results indicate that tyrosine at position 655 of the Omicron S protein is responsible for preferential usage of the cathepsin B/L-dependent endosomal entry pathway over the metalloproteinase-or TMPRSS2-dependent cell surface entry pathway.
To investigate whether the substitution of tyrosine for histidine at position 655 in the Omicron S protein enhances the cell surface pathways or impairs the endosome pathway, we analyzed the effect of the 655H substitution on infectivity through each pathway. Luciferase activities from HEC50B cells infected with Omicron and the (Omicron S +655H) pseudoviruses were almost the same (Fig. 3E). The Omicron S +655H pseudovirus infected cells more efficiently than the Omicron pseudovirus when the cathepsin B/L-dependent endosome pathway was blocked by E-64d (Fig. 3E). In contrast, the Omicron +655H pseudovirus was less infectious than the Omicron pseudovirus when the metalloproteinase pathway was blocked by marimastat (Fig.   3E). Moreover, the Omicron S +655H pseudovirus was more infectious to Calu-3 cells, which have only the TMPRSS2 pathway, and less infectious to OVISE cells, which have only the endosome pathway, than was the Omicron pseudovirus. Western blot analysis of the S proteins on the pseudoviruses revealed no significant difference in the cleavage status between Omicron S and Omicron S +655H (Fig. 3F). These results strongly suggest that the H655Y mutation of Omicron S is responsible for alteration of the usage of the entry pathways by facilitating entry through the cathepsin B/L-dependent endosome pathway and impairing entry through the metalloproteinase-and TMPRSS2-dependent cell surface pathways without affecting the S1/S2 cleavage status.

Discussion
We recently reported that host metalloproteinases are involved in cell surface fusion for SARS-CoV-2 entry, independent of TMPRSS2, and that the pathway SARS-CoV-2 uses for cell surface entry is cell type-dependent [21]. Although tissues and cells preferentially use the metalloproteinase pathway for SARS-CoV-2 infection in vivo are still unknown, reduced usage of the metalloproteinase pathway by Omicron could contribute to the modulation of tissue and cell tropism. Moreover, our data clearly show that Omicron displays a significantly reduced metalloproteinase-dependent fusogenic activity. Given that the fusogenicity of SARS-CoV-2 correlates well with disease severity, the low fusogenic activity of Omicron for metalloproteinase dependence is likely to contribute to the less severe disease phenotype of Omicron in cooperation with the low fusogenicity mediated by TMPRSS2. Conversely, the high fusogenic activity mediated by metalloproteinases, as observed in Delta (Fig. 2AB), may lead to severe pathogenicity of SARS-CoV-2. Thus, fusogenicity mediated by metalloproteinases, as well as TEMPRSS2, may be useful as a barometer for the pathological severity of SARS-CoV-2 variants.
In this study, we successfully identified a single mutation, H655Y, in the Omicron S protein, which is responsible for the altered entry properties. Furthermore, our data reveal that the H655Y mutation facilitates entry through the endosomal pathway and impaired entry through the cell surface pathways (Fig. 3E). H655Y is a common mutation that is present in BA.1, BA.2, and BA.3, variants classified as Omicron, and is considered to be an important mutation that defines the characteristics of Omicron. in the WT virus, the H655Y mutation has been reported in a variety of SARS-CoV-2 infection studies using cell culture and animal models [31][32][33][34][35]. Despite these facts, only a few studies have explored the impact of H655Y mutation on biological and pathological phenotypes. One research group reported that the H655Y mutation in the WT virus enhances viral replication and transmission, possibly through increased S1/S2 cleavage [35].
According to these results, the H655Y mutation in the WT S backbone appears to have different effects on viral phenotypes than that in the Omicron S backbone. In addition, the H665Y mutation was observed in the gamma (P.1) variant among VOCs. However, the role of this mutation in the phenotype of the gamma variant has not yet been investigated. Further studies are needed to assess how the H655Y mutation affects the biological and pathological phenotypes of WT virus and other VOCs than the Omicron.
Inhibition of S1/S2 cleavage has been reported to reduce the usage of the TMPRSS2 and metalloproteinase pathways and increases the usage of the endosomal pathway [15,21,30,36]. Furthermore, recent studies reported that the Omicron S protein expressed in some cell types, such as VeroE6-TMPRSS2 and HEK293 cells, is poorly cleaved into S1 and S2 compared to the WT S, resulting in a reduction in the fusogenic activity of Omicron [2,12,13]. However, we did not observe any significant differences in the cleavage of the S protein in the virions and infected HEC50B cells between Omicron S and WT S cells, even when the cell surface entry pathway and fusogenic activity were markedly suppressed ( Fig. 1D and 2C). These data strongly suggest that Omicron S has the ability to change the entry pathway usage and fusogenicity, regardless of the S1/S2 cleavage status, in some cell types used in this study. Importantly, pseudovirus analysis shows that the H655Y mutation of the Omicron S causes a change in entry pathway usage without affecting the S1/S2 cleavage status, strongly suggesting that H655Y regulates entry properties through mechanisms other than the reduction of the S1/S2 cleavage efficiency.
A putative mechanism is that the H655Y mutation alters the Omicron entry pathway by affecting the stability of the S1/S2 complex formed after cleavage by furin-like proteases. The H655Y mutation is localized in subdomain 2 (SD2) of the S1 domain. Structural studies of the SARS-CoV-2 S protein indicate that SD2 is proximal to the S2 domain [37][38][39], suggesting that the stability of the S1/S2 complex may be influenced by some amino acid mutations in SD2. In fact, the D614G mutation in SD2 destabilizes the interface between S1 and S2, facilitating disorder of a region including the S2' cleavage site, leading to increased infectivity [37,39]. In contrast, the H655Y mutation in SD2 may make the S1/S2 complex more stable in the backbone of the Omicron S. According to the recently reported three-dimensional structure of Omicron S [40], the S trimer has a structure that is broadly common to that of WT S (SFig. 5AB). The Cterminal peptide of the S1 domain containing the H655Y mutation interacts with the N-terminal peptide of the S2 domain. Comparing the structures around amino acid position 655 in the spike protein of Omicron and WT, both spikes have hydrogen bonds between the two main chains (position 655-656 in S1 and position 694-695 in S2). Interestingly, Omicron S has an additional hydrogen bond between the side chains of the tyrosine residue at position 655 (Y655) in S1 and the threonine residue at position 696 (T696) in S2) (SFig. 5C). This additional hydrogen bond in Omicron S suggests that it may stabilize the S1/S2 complex. The increased stability possibly prevents TMPRSS2-and metalloproteinase-mediated S2' priming on the cell surface by interrupting the exposure of the S2' priming site. However, increased stability may have a positive effect on the endosomal entry of Omicron. For instance, a stable S1/S2 complex may increase the chance of endocytosis of viral particles by persistent anchoring of viral particles to the target cell membrane. In other words, stabilization of the binding between the S1 domain bound to ACE2 and the S2 domain anchored to the viral membrane may allow for efficient internalization of viral particles during endocytosis, which involves a dynamic change in cell membrane curvature.
Elucidation of the differences between the Omicron S and the mutated S (Omicron +655H) by structural and biochemical analyses is needed to explain how the S H665Y mutation changes the relative usage of entry pathways in the Omicron variant.  Before the supernatants were harvested, the cells were incubated for an additional 48 h (for UT-NCGM02, QHN001, and TY11-927) or 72 h (for TY38-873). The virus titer was determined by a plaque assay using VeroE6-TMPRSS2 (JCRB 1819) cells.

Preparation of pseudotype VSV viral particles and infection experiments
VSV particles were prepared as previously described [21,42]. For the infection assay, the target

Quantification of intracellular SARS-CoV-2 RNA
Intracellular SARS-CoV-2 RNA was quantified as previously described [21]. Cell lysis and cDNA synthesis were performed using the SuperPrep II Cell Lysis & RT Kit (SCQ-401, TOYOBO, Osaka, Japan) following the manufacturer's instructions. Viral RNA was detected using THUNDERBIRD SYBR qPCR Mix (TOYOBO) at 95 °C for 3 min, followed by 50 cycles of 95 °C for 10 s and 60 °C for 1 min. Fluorescence was detected using a CFX Connect™ Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). The mRNA expression level of ribosomal protein L13a (Rpl13a) in each sample was used to standardize the data. The primer sequences used were 5'-AAATTTTGGGGACCAGGAAC-3' as forward primer and 5'-TGGCAGCTGTGTAGGTCAAC-3' as reverse primer for the SARS-CoV-2 N gene; 5'-TGTTTGACGGCATCCCAC-3' as forward primer and 5'-CTGTCACTGCCTGGTACTTC-3'as reverse primer for the human Rpl13a gene; and 5'-CTCAAGGTTGTGCGTCTGAA-3' as 23 forward primer and 5'-CTGTCACTGCCTGGTACTTCCA-3' as reverse primer for the African green monkey Rpl13a gene.

Statistical analysis
Statistically significant differences between mean values were determined using a two-tailed Student's t-test. Dunnett's and Tukey's tests were used for multiple comparisons. All data represent three independent experiments, and values represent the mean ± standard deviation, with p<0.05, considered statistically significant.