The polybasic cleavage site in the SARS-CoV-2 spike modulates viral sensitivity to Type I IFN and IFITM2

The cellular entry of severe acute respiratory syndrome-associated coronaviruses types 1 and 2 (SARS-CoV-1 and -2) requires sequential protease processing of the viral spike glycoprotein (S). The presence of a polybasic cleavage site in SARS-CoV-2 S at the S1/S2 boundary has been suggested to be a factor in the increased transmissibility of SARS-CoV-2 compared to SARS-CoV-1 by facilitating maturation of the S precursor by furin-like proteases in the producer cells rather than endosomal cathepsins in the target. We investigate the relevance of the polybasic cleavage site in the route of entry of SARS-CoV-2 and the consequences this has for sensitivity to interferons, and more specifically, the IFN-induced transmembrane (IFITM) protein family that inhibit entry of diverse enveloped viruses. We found that SARS-CoV-2 is restricted predominantly by IFITM2 and the degree of this restriction is governed by route of viral entry. Removal of the cleavage site in the spike protein renders SARS-CoV-2 entry highly pH- and cathepsin-dependent in late endosomes where, like SARS-CoV-1 S, it is more sensitive to IFITM2 restriction. Furthermore, we find that potent inhibition of SARS-CoV-2 replication by type I but not type II IFNs is alleviated by targeted depletion of IFITM2 expression. We propose that the polybasic cleavage site allows SARS-CoV-2 to mediate viral entry in a pH-independent manner, in part to mitigate against IFITM-mediated restriction and promote replication and transmission. This suggests therapeutic strategies that target furin-mediated cleavage of SARS-CoV-2 S may reduce viral replication through the activity of type I IFNs. IMPORTANCE The furin cleavage site in the S protein is a distinguishing feature of SARS-CoV-2 and has been proposed to be a determinant for the higher transmissibility between individuals compared to SARS-CoV-1. One explanation for this is that it permits more efficient activation of fusion at or near the cell surface rather than requiring processing in the endosome of the target cell. Here we show that SARS-CoV-2 is inhibited by antiviral membrane protein IFITM2, and that the sensitivity is exacerbated by deletion of the furin cleavage site which restricts viral entry to low pH compartments. Furthermore, we find that IFITM2 is a significant effector of the antiviral activity of type I interferons against SARS-CoV-2 replication. We suggest one role of the furin cleavage site is to reduce SARS-CoV-2 sensitivity to innate immune restriction, and thus may represent a potential therapeutic target for COVID-19 treatment development.


INTRODUCTION 69
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a novel 70 coronavirus that was identified in early 2020 (Wu et al., 2020). Entry of SARS-CoV-2 71 into the cell is initiated by the spike glycoprotein binding to its receptor, angiotensin-72 converting enzyme 2 (ACE2) (Hoffmann, Kleine-Weber, Schroeder, et al., 2020). 73 Spike is a type I transmembrane protein that is synthesised as a polyprotein 74 precursor and requires two steps of proteolytic cleavage at the S1/S2 boundary and 75 at the S2' site in order to mediate fusion of the viral and cell membranes. Due to the 76 insertion of four amino acids at the S1/S2 boundary of SARS-CoV-2 spike, with the 77 sequence 681PRRA|RSV687, SARS-CoV-2 spike contains a canonical furin-like  Nevertheless, similar mutations have only been found rarely in a small number of 89 the virus is highly IFN sensitive, particularly to IFNb and IFNg, indicating that a 143 number of ISGs have direct antiviral effects against SARS-CoV-2. 144 In order to address the activities of ISGs directed against spike-mediated 145 entry, we first determined whether we could recapitulate the IFN phenotypes 146 observed above using pseudotyped lentiviral vectors (PLVs). We generated PLVs 147 containing SARS-CoV-2 spike bearing a luciferase reporter gene and tested them for 148 sensitivity to IFNs on A549-ACE2. Similar to full-length SARS-CoV-2, we found that 149 PLVs with SARS-CoV-2 spike are also highly sensitive to IFNb and IFNg ( Figure 1C  To test the impact of each individual IFITM on SARS-CoV-2 infection, we 162 generated stable A549-ACE2 cell lines expressing each human antiviral IFITM 163 ( Figure 2A) and infected these with PLVs of SARS-CoV-2. We found that SARS-164 CoV-2 showed a small but significant sensitivity to IFITM1 and a comparatively 165 greater sensitivity to IFITM2 ( Figure 2B). We recapitulated these phenotypes by challenging the A549-ACE2-IFITM cells with SARS-CoV-2 at increasing MOIs and 167 using the supernatant of these cells 48 hours later to infect Vero-E6 cells, and 168 measuring viral infectivity by staining for N protein ( Figure 2C). At low MOI, SARS-169 CoV-2 was particularly sensitive to IFITM2 but not IFITM3, with an inhibitory effect 170 seen with IFITM1, and these sensitivities were ameliorated at high viral inputs.  Figure 2D) and infected these cells with both PLVs and replication competent 182 SARS-CoV-2 and found that infection was not inhibited, but rather was enhanced by 183 the presence of IFITM2 Y19F ( Figure 2E, 2F). Although surprising that 184 mislocalisation of IFITM2 resulted in enhancement of infection rather than simply an 185 absence of restriction, these data are consistent with a recent report suggesting that 186 similar mutants of IFITM3 enhance SARS-CoV-2 infection (Shi et al., 2020). These 187 data suggest that the localisation of IFITM2 to endosomes or its recruitment to 188 clathrin-coated pits at the plasma membrane is key to its inhibition of SARS-CoV-2 189 entry.

The polybasic cleavage site determines sensitivity to IFITM2 in the presence or 191 absence of TMPRSS2 192
A major difference between the spike protein of SARS-CoV-2 and the majority of the 193 phylogenetically related bat Sarbecoviruses, including SARS-CoV-1, is the presence 194 of the polybasic cleavage site at the S1/S2 boundary which facilitates the processing 195 of spike to S1/S2 during viral assembly in the producer cell rather than during entry 196 of the target ( Figure 3A). As this feature has been proposed to be associated with 197 the increased transmissibility of SARS-CoV-2, we hypothesized that it might affect 198 the sensitivity of the virus to IFITM2. To investigate this, we deleted the polybasic 199 cleavage site from SARS-CoV-2 (while preserving the adjacent RS cathepsin site) 200 and swapped the corresponding region from (P681-A684) SARS-CoV-2 into SARS-201 CoV-1, generating SARS-CoV-2DPRRA and SARS-CoV-1 PRRA respectively 202 ( Figure 3A). We made PLVs of these mutants and analysed spike expression and 203 virion incorporation by western blot using a polyclonal antibody against SARS-CoV-204 1/2 S2 ( Figure 3B). We found that all spike proteins were equivalently expressed in 205 the transfected producer 293T-17 cell. As expected, the SARS-CoV-1 spike exists 206 predominantly as the S1/2 precursor on pelleted virions in the supernatant. By 207 contrast, processed S2 was the predominant species found on virions pseudotyped 208 with SARS-CoV-2 spike, indicating furin-mediated cleavage during virion assembly. 209 By contrast, as expected SARS-CoV-2DPRRA was no longer cleaved. Insertion of 210 the SARS-CoV-2 cleavage site into SARS-CoV-1 was sufficient to lead to processed 211 spike, however this was not as efficient as in SARS-CoV-2, with virions incorporating 212 both cleaved and uncleaved spikes ( Figure 3B). In keeping with results from others, 213 in Vero-E6 cells SARS-CoV-2DPRRA PLVs had a markedly increased infectivity of 214 approximately 50-fold in A549-ACE2 cells compared to the wild type spike, Pöhlmann, 2020). Addition of the PRRA site to SARS-CoV-1 slightly reduced titres. 217 Since SARS-CoV-2 requires TMPRSS2 in the target cells to activate spike for entry, 218 we over-expressed TMPRSS2 in A549-ACE2 cells by retroviral transduction and 219 found that this specifically enhanced infection of SARS-CoV-2 PLVs, indicating that 220 in the absence of TMPRSS2 expression, much of the SARS-CoV-2 inoculum is not 221 infectious in these cells. 222 We then tested IFITM sensitivity of these PLVs in A549-ACE2 cells with and 223 without TMPRSS2 overexpression ( Figure 3D and 3E). As expected, SARS-CoV-2 224 was sensitive to both IFITM1 and IFITM2 in A549-ACE2 cells ( Fig 3D). SARS-CoV-1 225 PLVs were significantly more sensitive to IFITM2 but displayed no restriction by 226 IFITM1, suggestive of distinct subcellular site of entry between SARS-CoV-1 and 227 SARS-CoV-2. Interestingly deletion of PRRA in SARS-CoV-2 rendered this spike as 228 sensitive as SARS-CoV-1 to IFITM2 and slightly reduced the IFITM1 sensitivity. By 229 contrast the addition of a cleavage site to SARS-CoV-1 significantly reduced IFITM2 230 sensitivity, albeit not to the levels of the fully cleaved SARS-CoV-2 spike. When we 231 overexpressed TMPRSS2, we found that while IFITM1 sensitivity of SARS-CoV-2 232 could be abolished, this was not sufficient to rescue SARS-CoV-2 or SARS-CoV-233 2DPRRA from IFITM2. Thus, the presence of the polybasic cleavage site markedly 234 reduces the sensitivity of SARS CoV-2 S-mediated entry to IFITM2, suggesting it 235 affects the route of cellular entry into the cell and distinguishes it from SARS CoV-1. 236 To address the effects of spike cleavage on route of entry, we first determined 237 the pH-sensitivity of the spike cleavage mutants using concanamycin A (ConA), an 238 inhibitor of the vacuolar ATPase in late endosomes ( Figure 4A). As expected, SARS-239 CoV-1 PLVs were exquisitely sensitive (1000 fold) to ConA inhibition in A549-ACE2 240 indicating that entry was occurring exclusively in a low pH endosomal compartment. 241 In the presence of TMPRSS2, SARS-CoV-1 pH sensitivity was reduced, but entry 242 still remained 20-50 fold lower suggesting any enhanced S2' processing was not 243 sufficient to abolish pH-dependent entry. Similarly, while insertion of a partially 244 processed polybasic cleavage site in SARS-CoV-1 reduced but did not abolish pH-245 dependent entry in either cell type. By contrast, entry of SARS CoV-2 PLVs was only 246 mildly affected (2-3 fold) by ConA treatment irrespective of TMPRSS2 247 overexpression, indicating that most viral entry was occurring at neutral pH, and 248 TMPRSS2 was enhancing entry at this point rather than elsewhere in the cell. 249 Similar to SARS-CoV-1, deletion of the PRRA site from SARS-CoV-2 rendered PLVs 250 strictly pH-dependent without affecting titre. In keeping with these data, unlike SARS-251 CoV-2, SARS-CoV-1 and SARS-CoV-2DPRRA-mediated entry was inhibited by the 252 endosomal cathepsin inhibitor E64D but not the TMPRSS inhibitor Camostat ( Figure  253 4B-I). By contrast, SARS-CoV-2 only was sensitive to E64D in TMPRSS2-254 overexpressing cells. Together these data suggest that S1/S2 cleavage by furin in 255 the producer cell promotes TMPRSS2-mediated entry at the plasma membrane or 256 soon after internalization, and abolishes the requirement for cathepsin-mediated 257 processing in the acidic endosomal compartments. The data further suggest that in 258 the absence of abundant TMPRSS2 at the cell surface, the processed SARS-CoV-2 259 cannot efficiently enter through a low pH compartment. Thus, the PRRA site dictates 260 the route of entry into the cell and thus its sensitivity to IFITM proteins that occupy 261 different cellular locations. 262

IFITM2 contributes to the antiviral restriction of SARS-CoV-2 by IFNb 263
Having established that IFITM2 can restrict SARS-CoV-2 depending on its 264 mechanism of entry, we wanted to determine how much of the inhibition of replication-competent SARS-CoV-2 by IFNb and IFNg could be attributed to IFITM2. 266 We examined the expression of IFITM2 and IFITM3 in IFN-treated A549-ACE2 and 267 observed a robust upregulation of both IFITM2 and IFITM3 following treatment with 268 IFNb. By contrast, while IFITM3 was also robustly induced by IFNg, IFITM2 was 269 weakly induced ( Figure 5A). Using siRNAs against IFITM2 that rescued SARS-CoV-  Figure 5G). Thus IFITM2-mediated entry restriction is a major type-I 281 IFN activity that constitutes an antiviral state, blocking the replication of SARS CoV-282

DISCUSSION 284
In this study we have provided evidence that IFITM2 has potent inhibitory activity 285 against SARS-CoV-2 entry and constitutes at least part of the antiviral activity 286 conferred by treatment of target cells with IFNb. Furthermore, we find that the 287 presence or absence of the polybasic cleavage site, which facilitates pH-288 independent entry of SARS-CoV-2, modulates the sensitivity of the virus to IFITM2 289

In contrast to SARS-CoV-1 and other related SARS-like CoVs in bats, SARS-290
CoV-2 is distinguished by the presence of a furin cleavage site at the S1/S2 291 boundary. This leads to the spike on SARS-CoV-2 virions being predominantly 292 cleaved in the producer cell rather than by cathepsins during endocytic entry into the 293 target cell. This renders its entry pH-independent, suggesting fusion occurs at, or 294 near, the cell surface. Recent evidence further indicates that the furin cleavage 295 generates a C-terminal ligand on S1 that interacts with neuropilin-1 (NRP-1) on the can actively hinder infection. Herein we show that while wild-type spike-mediated 306 entry is insensitive to inhibition of endosomal pH, the cleavage mutant is strictly 307 dependent on endosome acidification and cathepsins. Interestingly, for efficient 308 entry, SARS-CoV-2 requires high TMPRSS2 expression to activate the fusion 309 mechanism by cleaving S2'. However, in cells where TMPRSS2 is limiting, SARS-310 CoV-1 and SARS-CoV-2DPRRA entry is far more efficient. Thus, in its uncleaved 311 form SARS-CoV-2 spike can mediate entry in endosomes, but in its mature form 312 entry cannot be rescued in low pH compartments of TMPRSS2-low cells. This would 313 imply that the cleaved spike is unstable at endosomal pH, and interestingly recent 314 studies from the Kwong group indicate that conformational dynamics of the RBD are 315 also pH sensitive (Zhou et al., 2020). Despite this potential greater fragility of the 316 SARS-CoV-2 trimer, the furin cleavage site appears to be essential for replication in 317 primary airway epithelium and for transmission in ferret models (Peacock et al., 318 2020). We suggest that one of the reasons pH-independent fusion at or near the cell 319 surface is maintained is to mitigate the antiviral activity of IFITM proteins, particularly 320

IFITM2. 321
The localization of IFITMs largely defines which viruses they can restrict. 322 to the plasma membrane and outside of clathrin coated pits by abolishing AP2 interaction, we see similar enhancement effects to those seen by the Yount group 365 with IFITM3 (Shi et al., 2020). Why this happens is not known, but given the effects 366 that IFITMs have on membrane fluidity, this may be an indirect effect on the surface 367 levels and distribution of entry cofactors at the plasma membrane. It also suggests 368 why there may be an association of the rs12252-C polymorphism that expresses an 369 N-terminally truncated IFITM3 with COVID-19 severity (Gómez et al., 2021). 370

Whilst they can be incorporated into nascent virion membranes and exert an antiviral
Restriction by IFITM2 but not IFITM3 is surprising. This would suggest that IFITM2 371 localization is not only limited to later endosomes than IFITM3, but may also reside 372 in distinct localizations at or near the plasma membrane dependent on its AP2-373 binding site. 374 In addition to examining the sensitivity of SARS-CoV-2 to individual IFITM 375 proteins, we also showed that IFITM2 knockdown is sufficient to alleviate much of 376 the antiviral effect of pretreating A549 cells with type I, but not type II IFN. Studies In A549-ACE2 cells, IFITM2 is more potently induced by IFNb than IFNg and 391 its knockdown substantially reduces the sensitivity of the virus to IFNb-induced 392 restriction. The sequence similarity between IFITM2 and IFITM3 means that it is 393 difficult to knockdown one without affecting the other. The lack of IFITM3 restriction 394 when expressed alone, and its potent expression after both IFNb and IFNg treatment 395 would argue against IFITM3 playing the major role. However, given that IFITMs can 396 interact with each other, we cannot rule out that IFITM1 or IFITM3 play a role in 397 potentiating IFITM2's antiviral activity after IFN induction. The former is a distinct 398 possibility as IFITM2 knockdown fully rescues SARS-CoV-2 from IFNb treatment in 399 cells over-expressing TMPRSS2. Since we found that the minor restriction conferred 400 by IFITM1 alone is abolished by TMPRSS2 expression, a plausible explanation is 401 that more robust S2' activation of SARS-CoV-2 spike at the cell surface overcomes 402 IFITM1inhibition by saturating its activity (Weston et al., 2016). While it is surprising 403 that IFNb has no effect in these cells when IFITM2 is knocked-down, we would 404 caution against interpreting that IFITM2 is the only ISG targeting SARS-CoV-2 405 replication. The rapidity and burst-size of SARS-CoV-2 replication in culture may 406 render other relevant antiviral proteins difficult to measure. Furthermore, the virus 407 encodes a number of antagonists of antiviral pathways (Xia et al., 2020). As shown 408 clearly by the IFNg phenotype, expression of other ISGs or their differential 409 regulation may make a give antiviral more less potent. Of note, the IFNg-mediated 410 inhibition of SARS-CoV-2 is in part mediated through the zinc-finger antiviral protein 411 (ZAP) (Nchioua et al., 2020). In sum we show that IFITM2 is a key antiviral protein targeting SARS-CoV-2 430 entry, and its activity is modulated by the furin-cleavage site in spike. These data 431 therefore suggest that therapeutic strategies which upregulate IFITM2 in epithelial 432 tissues or inhibit furin-mediated cleavage of spike may render the virus more 433 sensitive to innate-immune mediated control.