DC-SIGN and L-SIGN both serve as alternative receptors for SARS-CoV entry independently of ACE2.
To determine whether DC-SIGN or L-SIGN could serve as alternative receptors for SARS-CoV rather than simply as enhancer factors, infectivity of SARS pseudoviruses (murine leukemia virus pseudotyped with S glycoprotein) was examined by using HeLa cells transfected with plasmids encoding these proteins. As shown in Fig.
1A, HeLa cells transfected with pcDNA empty vector were completely refractory to SARS pseudovirus infection. In contrast, cells transfected with ACE2-expressing plasmid efficiently supported pseudovirus infection. HeLa cells expressing either DC-SIGN or L-SIGN also were susceptible to infection, albeit considerably less so than those expressing ACE2. Although the difference was minimal and not statistically significant, cells expressing L-SIGN were consistently more susceptible to infection than those expressing DC-SIGN.
We have previously described the development of an ACE2-derived peptide (P6) that potently inhibits SARS pseudovirus infections of Vero E6 or HeLa cells expressing ACE2 with a 50% inhibitory concentration of approximately 100 nM (
13). This peptide consists of two discontinuous segments of ACE2 (amino acid residues 22 to 44 and 351 to 357) artificially linked by a single glycine residue. These determinants have been shown to interact, biochemically and structurally, with the RBM of S protein and are critical for mediating SARS-CoV infection (
23,
25). To unequivocally demonstrate that pseudovirus infections of HeLa cells expressing L-SIGN are not mediated through ACE2, infectivity was examined in the presence of the P6 peptide. As shown in Fig.
1B, infection mediated by ACE2 was potently inhibited by P6 peptide in a dose-dependent manner as previously described (
13). In contrast, no inhibition was observed for L-SIGN-mediated infection, even in the presence of 100 μM. Similar results were observed for DC-SIGN-mediated infections (data not shown). These results not only indicate that DC/L-SIGN serve as alternative receptors but also indicate that the binding site(s) of these proteins on S protein is distinct from the site that ACE2 binds.
DC/L-SIGN are members of a C-type lectin family, the interactions of which with ligands are carbohydrate dependent (
2,
14,
26); they specifically recognize high-mannose glycans (
10). Mannan, a carbohydrate composed of high mannose, inhibits binding of ligands to DC/L-SIGN. To demonstrate that infections of HeLa cells expressing DC/L-SIGN by our SARS pseudoviruses are indeed mediated by DC/L-SIGN, infection of HeLa cells expressing either ACE2 or L-SIGN were carried out in the presence of various amounts of mannan (Fig.
1C). As expected, L-SIGN-mediated infections were inhibited by mannan in a dose-dependent manner. In contrast, ACE2-mediated infections were not affected by mannan.
To further characterize virus entry mediated by ACE2 and DC/L-SIGN, sensitivity of pseudoviruses to neutralizing MAbs was evaluated. Four MAbs (CDC-336, CDC-341, CDC-523, and CDC-540; obtained from Lia Haynes at the Centers for Disease Control and Prevention) were evaluated. These antibodies were generated from mice immunized with whole inactivated SARS-CoV particles (
44). The epitope recognized by CDC-341 is amino acid residues 490 to 510, which is at the C-terminal end of the RBD. The epitopes of the other MAbs have not yet been determined. Regardless, all four MAbs inhibited ACE2-mediated virus entry (Fig.
1D). This is not surprising, since they were screened for their ability to block virus infection of Vero E6 cells. In contrast, none of the MAbs inhibited L-SIGN-mediated virus entry. This is not because L-SIGN-mediated infections are intrinsically difficult to inhibit, since polyclonal antiserum from mice immunized with
Lactobacillus casei expressing S-protein fragments (
22) was able to inhibit infections mediated by both receptors. Together, these results demonstrate that DC/L-SIGN can mediate SARS-CoV infections independently of ACE2.
Glycans on specific sites are critical for L-SIGN-mediated infections.
Treatment of glycoproteins with glycosidase allows only gross assessment of the potential importance of carbohydrate moieties in protein function, because glycans are removed indiscriminately. In addition, the removal of a large mass of glycans could alter global conformation of the protein structure. In this regard, characterizing S proteins with individual glycosylation sites eliminated by site-directed mutagenesis could provide more accurate information on the functional importance of glycans in virus entry. In particular, we were interested in (i) whether glycans at certain glycosylation sites are more important than others for virus entry and (ii) whether glycans at different sites have different roles in ACE2- and DC/L-SIGN-mediated infections.
There are 23 potential asparagine (N)-linked glycosylation sites on S protein (Fig.
4A). On a linear map of S glycoprotein, these sites appear to be distributed into three distinct clusters: cluster I at the N terminus (aa 29, 65, 73, 109, 118, 119, 158, 227, 269, 318, 330, and 357), cluster II in the middle of the protein near the border between S1- and S2-like domains (aa 589, 602, 691, 699, and 783), and cluster III at the C terminus (aa 1056, 1080, 1116, 1140, 1155, and 1176). To date, glycosylation at 13 of these sites (aa 118, 119, 227, 269, 318, 330, 357, 783, 1056, 1080, 1140, 1155, and 1176) have been confirmed by either mass spectrometric (
19,
49) or biochemical (
6) analyses. The glycosylation status of other sites needs to be further determined.
Since the S1 domain (aa 1 to ∼680) (
27,
41) is responsible for binding ACE2, we focused on characterizing 12 N-linked glycosylation sites in cluster I. Asparagine residues of the canonical NXS/T motif were individually mutated to glutamine (Q), which differs by only a single methylene group and represents the most conservative amino acid substitution. Twelve mutant SARS pseudoviruses were generated, and their infectivity in ACE2- or L-SIGN-expressing HeLa cells was compared to that of the wild-type pseudovirus. As shown in Fig.
4B, all of the mutant pseudoviruses exhibited near-wild-type levels of infectivity in ACE2-expressing cells. In contrast, four mutants exhibited marked defects in their ability to use L-SIGN (mutants N109Q, N118Q, N119Q, and N227Q). Their infectivity was only about 30 to 40% of that of the wild-type pseudovirus. This reduction is significant, considering that the infectivity of pseudoviruses treated with Endo H was approximately 30% of that of the untreated virus (Fig.
3B). A modest, but reproducible, reduction in infectivity also was observed for mutant N158Q. The loss of infectivity by these five mutant pseudoviruses is most likely due to a reduced ability of mutant S proteins to specifically interact with L-SIGN rather than gross misfolding of the protein, since ACE2-mediated virus infection is virtually unaffected. Not surprisingly, all of the mutant S proteins were expressed normally as demonstrated by Western immunoblotting (Fig.
4C).
To further characterize five mutant S proteins, kinetic parameters of pseudovirus infectivity were evaluated. To do so, pseudovirus infectivity was reevaluated using a 20-min adsorption period (the time viruses are allowed to adsorb onto cells before the inoculum is removed) in addition to the typical 60 min. This was done because we have previously shown that ACE2-mediated infectivity begins to plateau by 60 min and that phenotypic differences between the wild-type and mutant ACE2 proteins were better observed using a 20-min period (
13). The wild type and two mutants that did not exhibit significant defects (N29Q and N269Q) were examined as controls. As shown in Fig.
5A, the infectivity level of the wild-type pseudovirus at 20 min was about 70% of that at 60 min. The N29Q mutant pseudovirus exhibited an almost identical pattern. Although the level of infectivity of the N269Q mutant was slightly lower at 20 min, it increased to the wild-type level by 60 min. In contrast, the infectivity of the five other mutants was significantly lower than that of the wild type at 60 min. Interestingly, the infectivity of these mutants at 60 min was virtually identical to that at 20 min.
To better understand this unexpected observation, more detailed ACE2- and L-SIGN-mediated infection kinetic analyses were performed for mutants N109Q and N118Q. As shown in Fig.
5B (left panel), both mutants exhibited infection kinetics similar to those of the wild type when ACE2 was used. However, when L-SIGN was used, the mutants exhibited markedly different kinetics (Fig.
5B, right panel). In contrast to the wild type, the infectivity of which continued to increase beyond 20 min, infectivity levels of mutant viruses reached a plateau in 20 min, consistent with the results shown in Fig.
5A. The precise reason for this observation is not yet clear. It is also interesting that the initial kinetic of infection mediated by L-SIGN appeared faster than that mediated by ACE2. While infectivity of the wild-type pseudovirus reached 35% using L-SIGN within 5 min of incubation, ACE2-mediated infection reached only about 5%. This suggests that interactions between L-SIGN and glycans are less specific, and therefore take less time, than those between ACE2 and the RBD of S protein.
Thus far, the results of our study demonstrated that glycans are important for DC/L-SIGN-mediated SARS-CoV entry and that 5 of 12 glycosylation sites in cluster I are critical (i.e., N109, N118, N119, N158, and N227). Although mutating each of these glycosylation sites resulted in substantial reduction in L-SIGN-mediated infectivity, none of them was completely disruptive. One likely explanation for this observation is that utilization of L-SIGN per se does not require the presence of all five glycosylation sites. However, efficient utilization may require a critical density of glycans and needs all five sites. Thus, eliminating any one of these sites would result in a reduction, but not complete loss, of infectivity.
We next asked whether efficiency in utilizing L-SIGN correlated with the available number of glycosylation sites. One could hypothesize that eliminating a greater number of glycosylation sites simultaneously would render the protein progressively less efficient in utilizing L-SIGN. To test this hypothesis, we generated nine additional S proteins with multiple mutations in various combinations (double, triple, quadruple, or pentuple) and evaluated the infectivity of pseudoviruses using ACE2 or L-SIGN. Not unexpectedly, combining N29Q, N65Q, and N73Q mutations (i.e., mutant μ2-1 or μ3-1), none of which affected virus infectivity individually, reduced neither ACE2- nor L-SIGN-mediated infectivity (Fig.
6A). In contrast, we were surprised to observe that combining individual mutations that reduced virus infectivity (i.e., N109Q, N118Q, N119Q, N158Q, and N227Q) did not significantly worsen the effect, even when all five sites were mutated simultaneously (i.e., mutant μ5). None of these mutants exhibited significant reduction in infectivity using ACE2, except for that of μ5, which was at about 70% of the wild-type level. The expression level of μ5 mutant S protein was not significantly different from that of μ4 or the wild-type proteins (Fig.
6B), suggesting that this reduction in ACE2-mediated infectivity is not due to a defect in protein expression. Instead, mutating five glycosylation sites all at once most likely partially altered the conformation of the ACE2-binding domain.
Incomplete loss of infectivity by the μ5 mutant could be explained if glycosylation sites outside of cluster I also could be utilized for DC/L-SIGN-mediated virus entry (Fig.
4A). The obvious alternative sites could be those within cluster II. To evaluate the potential role of glycosylation sites in cluster II, five additional mutant S proteins were generated (Fig.
7). Only two of the five proteins (N589Q and N699Q) exhibited partial loss (about 60% of the wild-type level) in L-SIGN-mediated infectivity. All of the mutants exhibited normal infectivity using ACE2 as a receptor. These results indicate that specific glycosylation sites in both clusters I and II play a role in DC/L-SIGN-mediated virus infections.
Despite seemingly less efficient utilization of DC-SIGN by SARS-CoV pseudoviruses compared to that of L-SIGN, we nonetheless examined the effects of N-linked glycosylation site mutations on DC-SIGN-mediated pseudovirus infectivity. All seven glycosylation site mutations that affected L-SIGN usage were evaluated (i.e., N109Q, N118Q, N119Q, N158Q, N227Q, N589Q, and N699Q). As shown in Fig.
8, DC-SIGN-mediated infections were affected in a manner similar to that of L-SIGN-mediated infections. As expected, no significant loss of infectivity was observed for N330Q, N357Q, or the μ2-1 double mutant (N65Q/N73Q). These results suggest that DC-SIGN- and L-SIGN-mediated infections likely proceed by a similar mechanism.