SARS-CoV-2 Spike evolution influences GBP and IFITM sensitivity

SARS-CoV-2 spike requires proteolytic processing for viral entry. The presence of a polybasic furin-cleavage site (FCS) in spike, and evolution towards an optimised FCS by dominant variants of concern (VOCs), are linked to enhanced infectivity and transmission. Here we show that interferon-inducible antiviral restriction factors Guanylate binding proteins (GBP) 2 and 5 interfere with furin-mediated cleavage of SARS-CoV-2 spike and inhibit the infectivity of early-lineage Wuhan-Hu-1, while VOCs Alpha and Delta have evolved to escape restriction. Strikingly, we find Omicron is unique amongst VOCs, being restricted by GBP2/5, and also IFITM1, 2 and 3. Replacing the spike S2 domain in Omicron with Delta shows S2 is the determinant of entry route and IFITM sensitivity. We conclude that VOC evolution under different selective pressures has influenced sensitivity to spike-targeting restriction factors, with Omicron selecting spike changes that not only mediate antibody escape, and altered tropism, but also sensitivity to innate immunity.


31
SARS-CoV-2 infects cells by binding of the viral spike (S) protein to the angiotensin 32 converting enzyme 2 (ACE2) receptor on host cells 1 . For fusion to proceed after ACE2 33 binding, spike must be cleaved by host cell proteases to become activated and fusion-34 competent. In this step-wise process spike is pre-processed by furin-like proteases in virus-35 producing cells at the S1/S2 junction 2, 3 , followed by a second cleavage event at the S2' site To define mechanism, we analysed the effects of GBPs on spike S1/S2 cleavage by western 151 blotting. Visualising uncleaved (S) and cleaved spike (S2, S2/S ratio) revealed clear 152 differences in processing of Wuhan-Hu-1 spike on PV particles and in cell lysates in the 153 presence of GBP2 and 5, compared to either no GBPs, or GBP2 C588A and GBP5 C583A 154 mutants ( Fig. 2A, B). Quantifying this across all spike variants showed that GBP5 155 significantly reduced the amount of cleaved spike (lower S2/S ratio) in Wuhan-Hu-1, Wuhan-  167 168 Next, we tested whether decreasing spike expression can sensitise Wuhan-Hu-1 D614G,

169
Alpha and Delta PV to GBP2/5 restriction. Indeed, lowering the expression of spike by 170 transfecting decreasing amounts of spike plasmid during PV production restored the ability 171 of GBP2/5 to restrict Wuhan-Hu-1 D614G, and to a lesser extent Alpha and Delta infectivity 172 to levels seen for Wuhan-Hu-1 (Extended data Fig. 5A-D), whereas Omicron spike was even 173 further inhibited by GBP2/5 (Extended data Fig. 5E). Similar results were obtained when 174 measuring PV infectivity on HeLa-ACE2 cells, although Omicron again behaved as an outlier 175 (Extended data Fig. 5F-J). To determine whether increasing spike incorporation into PV 176 could render Wuhan-Hu-1 resistant to GBP5 restriction, we took an advantage of the fact 177 that SARS-CoV-2 spike contains an endoplasmic reticulum retention signal in the 178 cytoplasmic tail (CT) and that truncation of the last 19 residues of the spike cytoplasmic tail 179 (CT) removes this motif, boosting spike incorporation and particle infectivity during 180 pseudotyping 37, 38 . Concordantly, truncation of the Wuhan-Hu-1 spike CT (Wuhan CT) 181 enhanced PV infectivity 10-fold compared to PV containing the full-length spike (Extended 182 data Fig. 6A). However, both CT and full-length spike remained sensitive to GBP5 183 restriction (Extended data Fig. 6B), although CT spike was less sensitive compared to its 184 both the increase in incorporation of CT spike into PV, and GBP5 inhibition of CT and full-186 length spike cleavage (Extended data Fig. 6C). Taken together these data identify GBP2 187 and 5 as restriction factors that target SARS-CoV-2 by interfering with spike processing at 188 the S1/S2 cleavage site. Consequently, the infectivity of early-lineage variants (Wuhan-Hu- for Omicron compared to other VOCs (Fig. 3E). Efficient S1/S2 cleavage of Omicron spike is 210 consistent with this VOC containing the FCS optimising mutations P681H, N679K and 211 H655Y (Fig. 1A) that can enhance S1/S2 cleavage 14,15 . These data suggest that Omicron's 212 shift in cellular tropism is unlikely to be explained by inefficient spike cleavage in virions.

213
Measuring the sensitivity of PV made in the presence of GBPs to entry inhibitors showed 214 that GBP2 and 5 partially, but significantly, sensitised Wuhan-Hu-1 PV to inhibition by E64d,

241
Testing the effects of GBP5 expressed in PV producer cells, on the sensitivity of PV to IFITM 242 restriction in target cells, we found no significant difference in IFITM1 inhibition of Wuhan-

243
Hu-1 PV made in the presence or absence of GBP5 (Fig. 4N). For IFITM2/3, although PV 244 made in the presence of GBP5 appeared to show enhanced sensitivity to endosomal 245 IFITM2/3, this can be accounted for by the inhibitory effect of GBP5, evidenced by 246 comparing infectivity of GBP5 plus IFITM2 or 3, to GBP5 alone (Fig. 4N). Moreover, Wuhan-

247
Hu-1 D614G, Alpha and Delta PV sensitivity to IFITMs was similarly unaffected by GBPs 248 . Given that GBPs did not completely prevent S1/S2 spike cleavage in our 249 pseudovirus assay, this result is not unexpected. Similar results were obtained using live 250 viruses (VIC, Alpha and Delta) produced in presence of IFNγ, where inducing expression of 251 GBP2/5 did not sensitise virus to enhanced IFITM restriction (Fig. S9). Collectively these 252 data show that while combining GBPs and IFITMs can lead to an overall potent reduction of 253 infection, inhibiting furin cleavage by GBPs does not make a virus that is resistant to IFITMs 254 become sensitive to restriction.

256
Omicron virus is restricted by IFITM 1, 2 and 3 257 tested Omicron's sensitivity to IFITMs. Strikingly, we found Omicron was unique amongst 260 SARS-CoV-2 viruses we tested, being sensitive to inhibition of infection by IFITM1, 2 and 3 261 in Caco2 cells (Fig. 4G). Similar restriction of Omicron by IFITM1, 2 and 3 was seen in PV 262 assays (Fig. 4L). Next, we investigated the effects of combining GBP and IFITMs on 263 Omicron infection. Fig. 4R shows that Omicron PV made in the presence of GBP5 was not 264 significantly more sensitive to IFITM restriction. Similarly, IFNγ treatment in producer cells 265 did not further sensitise Omicron to IFITM inhibition (Extended data Fig. 9D) To test whether 266 increased spike incorporation changes Omicron sensitivity to GBP or IFITM restriction, we 267 used full-length and CT spike mutants (described in Extended data Fig. 6). As expected, 268 increasing spike incorporation into Omicron PV by truncation of the spike CT (Omicron CT) 269 enhanced PV infectivity 70-fold compared to PV containing the full-length Omicron spike.  repeat domain 1 (HR1) of spike that mediates viral fusion, as well as N856K adjacent to the 284 fusion peptide (Fig. 1A), which we hypothesised may play a role in Omicron's altered entry 285 and sensitivity to spike targeting restriction factors. Since Delta spike has the highest 286 fusogenicity 41 we swapped S2' domain of Omicron spike with that of Delta to produce 287 pseudoparticles with Omicron-S2'Delta spike chimera (Fig. 5H). PV containing the chimeric 288 Omicron-S2'Delta spike PV retained low particle infectivity (Fig. 5H), and remained sensitive to 289 GBP5 inhibition of spike cleavage, behaving like WT Omicron and not like Delta ( Fig. 5I

302
Innate immunity is a potent first-line host cell defense against viruses, upregulating a group 303 of interferon-stimulated genes (ISGs) which can act directly as restriction factors, targeting 304 key steps in viral replication and collectively inducing an antiviral state. SARS-CoV-2 triggers 305 innate immune sensing and induces an interferon-response 11, 27, 42 , upregulating canonical 306 ISGs including GBP2 and 5 in primary human airway epithelial cells 21 . Evolution of 307 mutations outside of spike allow for SARS-CoV-2 evasion/antagonism of innate immune 308 sensing 11 ; however, spike itself can be targeted by innate immune responses. Here we 309 discover that the interferon-inducible restriction factors GBP2 and 5 interfere with furin-310 mediated cleavage of SARS-CoV-2 spike and potently inhibit infection by the early-lineage 311 SARS-CoV-2 strain Wuhan-Hu-1 but that dominant VOCs have evolved differential 312 sensitivity to restriction. Using spikes from other coronaviruses with pandemic potential,

313
MERS-CoV and SARS-CoV-1, we confirm that restriction by GBP2/5 correlates with the 314 requirement for furin-mediated spike processing, such that MERS-CoV spike is sensitive, but 315 SARS-CoV-1 resistant, to inhibition by GBPs.

321
Indeed, we found that both Alpha and Delta were resistant to GBP-mediated inhibition of PV 322 and live virus infectivity. However, escape from restriction by these VOCs cannot be simply 323 attributed to an optimised FCS overcoming GBP inhibition of spike cleavage. This is 324 because both Alpha and Delta displayed the same GBP-mediated spike cleavage defect as 325 Wuhan-Hu-1. Moreover, the amount of cleaved spike incorporated into PV was comparable 326 between Wuhan-Hu-1 (restricted) and Alpha and Delta (not-restricted). In addition, Omicron 327 PV and virus infectivity was potently inhibited by GBP2/5, despite containing similar FCS-328 optimising mutations as Alpha/Delta. Instead, we argue that evolution of Alpha and Delta for 329 enhanced activity of spike, including better fusogenicity and infectivity 6,7,9,41,45 , are likely 330 inhibition. Therefore, we propose that it is the combined effects of mutations that have 336 accumulated in Alpha and Delta spikes, changing spike activity and infectivity, that have led 337 to evolution of these VOCs to escape inhibition by GBPs.

339
IFITMs are another family of ISGs that can target SARS-CoV-2 spike and restrict infection,

362
Our most striking observation is that Omicron is unique amongst the VOCs we tested in 363 being sensitive to inhibition by GBP2/5, and also IFITM1, 2 and 3. Omicron is also unique 364 amongst SARS-CoV-2 in its entry pathway, having evolving towards TMPRSS2-365 independence and endosomal-dependent fusion 30,31,32 . In agreement with these recent 366 reports, we show Omicron is significantly less infectious than other SARS-CoV-2 isolates on 367 while being more sensitive to the cathepsin inhibitor E64d, and now map this entry IFITM2 and 3, which are predominantly located within endosomes. Omicron contains the 372 same P681H FCS-optimising mutation as Alpha, as well as addition H655Y and N679K that 373 may further optimise the FCS 14, 15 . Indeed, we show that Omicron S1/S2 cleavage is notably 374 more efficient that either Alpha or Delta, with Omicron virus incorporating very little, if any, 375 uncleaved spike into virions. Therefore, we expected that Omicron would also be resistant to 376 GBPs, like other VOCs; however, Omicron was strikingly sensitive GBP2/5 with infectivity 377 inhibited to levels seen for Wuhan-Hu-1. Moreover, we were unable to rescue Omicron from 378 GBP or IFITM restriction by increasing the amount of incorporated spike into PVs, to a level 379 that increased the baseline infectivity by 70-fold. We therefore conclude that it is the inherent 380 differences Omicron's spike activity, rather than differences in S1/S2 cleavage efficiency or 381 spike incorporation, that make this VOC sensitive to GBP-mediated restriction of infection.

382
Omicron spike is reportedly less fusogenic 30, 31, 32 and compared to Wuhan-Hu-1, Omicron 383 contains 4 unique mutations in the S2' domain that mediates viral fusion ( Fig. 1A)     Omicron has taken a different evolutionary path to preceding VOCs. We propose a scenario 418 in which evolution of Omicron spike for neutralising antibody escape has influenced the 419 ability to evade innate immunity. The critical balance between viral evasion of innate and 420 adaptive immunity has precedent. This is borne out of studies of HIV-1 evolution in a host,

421
where HIV-1 isolates from early in infection (so called transmitter/founder viruses) are 422 completely resistant to IFITM restriction, but overtime, the selective pressure from adaptive 423 immunity, and the resulting neutralising antibody escape mutations in HIV-1 Env, leads to 424 viral isolates having increased sensitivity to IFITMs and interferons 47 . We propose that 425 similar processes have occurred during SARS-CoV-2 evolution to host, in which the need to 426 escape from neutralising antibody became the dominant selective pressure on Omicron, 427 resulting in a compensatory, but tolerable, increase in sensitivity to innate immunity, while 428 also impacting on spike activity and cell tropism. We predict that this interplay between 429 evasion of innate and adaptive immunity, and the consequences for transmission and 430 tropism, will be features of future SARS-CoV-2 evolution, and emergence of new VOCs, and 431 that linking this evolution to phenotype will become important aspects for understanding and 432 predicting SARS-CoV-2 biology, and ultimately pathogenesis.

481
Omicron BA.1 vector sequence excluding S2' domain was amplified using following primers: The fragments were assembled using NEBuilder HiFi DNA Assembly Cloning Kit (New