key: cord-0848959-eecwjrdc
authors: Conceicao, Carina; Thakur, Nazia; Human, Stacey; Kelly, James T.; Logan, Leanne; Bialy, Dagmara; Bhat, Sushant; Stevenson-Leggett, Phoebe; Zagrajek, Adrian K.; Hollinghurst, Philippa; Varga, Michal; Tsirigoti, Christina; Hammond, John A.; Maier, Helena J.; Bickerton, Erica; Shelton, Holly; Dietrich, Isabelle; Graham, Stephen C.; Bailey, Dalan
title: The SARS-CoV-2 Spike protein has a broad tropism for mammalian ACE2 proteins
date: 2020-06-18
journal: bioRxiv
DOI: 10.1101/2020.06.17.156471
sha: 92afe0816ddc7cf50fbb5d39931ddc5cb010201d
doc_id: 848959
cord_uid: eecwjrdc

SARS-CoV-2 emerged in late 2019, leading to the COVID-19 pandemic that continues to cause significant global mortality in human populations. Given its sequence similarity to SARS-CoV, as well as related coronaviruses circulating in bats, SARS-CoV-2 is thought to have originated in Chiroptera species in China. However, whether the virus spread directly to humans or through an intermediate host is currently unclear, as is the potential for this virus to infect companion animals, livestock and wildlife that could act as viral reservoirs. Using a combination of surrogate entry assays and live virus we demonstrate that, in addition to human ACE2, the Spike glycoprotein of SARS-CoV-2 has a broad host tropism for mammalian ACE2 receptors, despite divergence in the amino acids at the Spike receptor binding site on these proteins. Of the twenty-two different hosts we investigated, ACE2 proteins from dog, cat and rabbit were the most permissive to SARS-CoV-2, while bat and bird ACE2 proteins were the least efficiently used receptors. The absence of a significant tropism for any of the three genetically distinct bat ACE2 proteins we examined indicates that SARS-CoV-2 receptor usage likely shifted during zoonotic transmission from bats into people, possibly in an intermediate reservoir. Interestingly, while SARS-CoV-2 pseudoparticle entry was inefficient in cells bearing the ACE2 receptor from bats or birds the live virus was still able to enter these cells, albeit with markedly lower efficiency. The apparently broad tropism of SARS-CoV-2 at the point of viral entry confirms the potential risk of infection to a wide range of companion animals, livestock and wildlife.

particle entry in pseudotype assays, e.g. chicken ACE2, were still able to support live virus 97 entry at a high multiplicity of infection. This research has identified vertebrate species where 98 cell entry is most efficient, allowing prioritisation of in vivo challenge studies to assess disease 99 susceptibility. Combining this with increased surveillance and improved molecular diagnostics 100 could help to prevent future reverse zoonoses. 101 102 103

Results 104 105

The SARS-CoV-2 binding site on ACE2 is highly variable 106 107

Recent structural and functional data have shown that SARS-CoV, SARS-CoV-2 and other β-108 coronavirus (lineage B clade 1) Spike proteins bind the same domain in ACE2 to initiate viral 109 entry [5, 6, [8] [9] [10] . We thus hypothesised that SARS-CoV-2 could use the ACE2 receptor to 110 infect a range of non-human, non-bat hosts. To this end we synthesised expression constructs 111

for human ACE2 as well as orthologues from 22 other vertebrate species, including nine 112 companion animals (dogs, cats, rabbits, guinea pigs, hamsters, horses, rats, ferrets, 113 chinchilla), seven livestock species (chickens, cattle, sheep, goats, pigs, turkeys, buffalo) , four 114

bat species (horseshoe bat, fruit bat, little brown bat and flying fox bat), and two species 115

confirmed or suspected to be associated with previous coronavirus outbreaks (civet and 116 pangolin). There is 62 to 99% sequence identity between these proteins at the amino acid 117 level (76-99% when excluding the two bird sequences) and their phylogenetic relationships 118 are largely consistent with vertebrate phylogeny, although the guinea pig sequence was more 119 divergent than predicted (Fig.1A) . Examining the conservation of amino acids at the SARS-120

CoV-2 binding site on the surface of the ACE2 protein revealed a high degree of variation 121 across mammalian taxa (Fig.1B,C) , suggesting that SARS-CoV-2 receptor binding may vary 122 between potential hosts. This variation was also evident when aligning the 23 ACE2 123 sequences included in our study, which identified a number of highly variable residues within 124 the overlapping SARS-CoV and SARS-CoV-2 binding sites, including Q24, D30, K31, H34, 125 L79 and G354 (Fig.1D ). Our first step was to ensure efficient and equivalent surface 126 expression of these ACE2 proteins on target cells. To this end their N-terminal signal peptides 127

were replaced with a single sequence from the commercially available pDISPLAY construct 128 (Fig.1E ). In addition, the ectodomain was fused with a HA-epitope tag to allow the specific were barely detectable. The cause of this poor expression is unknown, potentially arising due 135

to errors in the ACE2 sequences available for these species. Since the available sequence 136 accuracy for these two genes would need to be explored further these two ACE2 proteins 137

were excluded from our subsequent experiments. 138 139

Receptor screening using surrogate entry assays identifies SARS-CoV-2 Spike as a 140

pan-tropic viral attachment protein 141 142

To examine the capacity of SARS-CoV-2 to enter cells bearing different ACE2 proteins we 143

used two related approaches. The first, based on the widely employed pseudotyping of 144 lentiviral particles with SARS-CoV-2 Spike [9], mimics particle entry. The second approach, 145

based on a quantitative cell-cell fusion assay we routinely employ for the morbilliviruses [14] , 146 assesses the capacity of Spike to induce cell-cell fusion following receptor engagement. In 147 both assays we used a codon-optimised SARS-CoV-2 Spike expression construct as the 148 fusogen, demonstrating robust and sensitive detection of either entry or fusion above 149 background (Sup. Fig.2A,B) . Supportive of our technical approach, replacing the human 150

ACE2 signal peptide with that found in pDISPLAY had no effect on pseudotype entry or cell-151 cell fusion (Sup. Fig.2 ). In addition, SARS-CoV-2 entry was shown only with human ACE2, but 152 not aminopeptidase N (APN) or dipeptidyl peptidase 4 (DPP4), the β-coronavirus group I and 153

MERS-CoV receptors, respectively (Sup. Fig.2) , indicating high specificity for both assays. 154

Using the classical pseudotype approach, which models particle engagement with receptors 155 on the surface of target cells, we demonstrated that SARS-CoV-2 Spike has a relatively broad 156 tropism for mammalian ACE2 receptors. Indeed, we observed that pangolin, dog, cat, horse, 157

sheep and water buffalo all sustained higher levels of entry than was seen with an equivalent 158 human ACE2 construct ( Fig.2A ; left heatmap, first column). In contrast, all three bat ACE2 159

proteins we analysed (fruit bat, little brown bat and horseshoe bat) sustained lower levels of 160 fusion than was seen with human ACE2, as did turkey and chicken ACE2, the only non-161 mammalian proteins tested. In accordance with previously published data on SARS-CoV and 162

SARS-CoV-2 usage of rodent ACE2 [1, 15] , rat ACE2 did not efficiently support SARS-CoV-2 163 particle entry. However, we observed that the ACE2 from hamsters did support pseudoparticle 164 entry, albeit less efficiently than human ACE2. 165

In the separate cell-cell fusion assay, which provides both luminescence and fluorescence-166 based monitoring of syncytia formation, a similar trend was observed with expression of 167 chinchilla, rabbit, hamster, pangolin, dog, cat, horse, pig, sheep, goat, water buffalo and cattle 168 ACE2 proteins on target cells all yielding higher signals than target cells expressing human 169 ACE2 ( Fig.2A ; left heatmap, second column). Similar to the pseudotype assay, expression of 170 all three bat ACE2 proteins resulted in less cell-cell fusion than that seen with human ACE2.

Example micrographs of GFP-positive SARS-CoV-2 Spike-induced syncytia are provided in 172 Sup. Fig.3 . The heatmaps presented in Fig.2A represent the average results from three 173 independent pseudotype and cell-cell assay receptor usage screens (with representative data 174 sets shown in Sup. Fig.4) . 175

Combining the results from all six screens demonstrates a significant degree of concordance 176 between the two experimental approaches. The only marked outlier is rabbit ACE2, which 177

repeatedly generated higher signals relative to human ACE2 in the cell-cell fusion assay 178 (Fig.2B ). Although the high correlation (Pearson r=0.73) was unsurprising, given that both 179 approaches rely on the same Spike-ACE2 engagement, fusogen activation and membrane 180 fusion process (albeit at virus-cell or cell-cell interfaces), there were some marked differences 181 in sensitivity. For the pseudotype system there was little appreciable evidence for particle entry 182 above background levels with ferret, rat, chicken, turkey or horseshoe bat ACE2, either in 183 vector control (pDISPLAY) transfected cells ( Fig.2A ; bottom row) or in ACE2-transfected cells 184

infected with a 'no glycoprotein' pseudoparticle control, NE (Sup. Fig.4 ). However, in the cell-185

cell system all of these receptors permitted Spike-mediated fusion, above the background 186 levels seen in pDISPLAY transfected cells ( Fig.2A ) or in effector cells not expressing SARS-187

CoV-2 Spike (Sup. Fig.4 ; No Spike), albeit at levels significantly lower than that seen for 188 human ACE2. This suggests that these receptors, whose structures are clearly not optimal for 189 SARS-CoV-2 entry, are still bound by the Spike protein. To facilitate comparison with existing data for SARS-CoV, we performed all the above 191 experiments side-by-side with SARS-CoV pseudotype and cell-cell assays (Fig.2A, right 192 heatmap and Sup. Fig.2,4) . While the receptor usage profile of SARS-CoV correlates 193 significantly with SARS-CoV-2, both in terms of pseudotype entry (Sup. Fig.5A ; r=0.86) and 194

cell-cell fusion (Sup. Fig.5B ; r=0.78), there were interesting divergences. In general, for CoV there was a better correlation between pseudotype entry and cell-cell fusion (Fig.2B, C; 196 Pearson r=0.73 [SARS-CoV-2] versus r=0.90 [SARS-CoV]), with no obvious outliers and less 197 variation between the two assays when examining receptors with low levels of associated 198

fusion, e.g. horseshoe bat ACE2 ( Fig.2A) . These differences may be due to the differing levels 199

of fusion seen with both viruses as well as the methodological approach taken. In our 200 experiments SARS-CoV-2 Spike is demonstrably more fusogenic than SARS-CoV, possibly 201 due to the presence of a furin-cleavage site between S1 and S2 [16] . Alongside a similar 202 restriction for bird and bat ACE2 proteins, our side-by-side comparison also identified 203

instances of varying restriction, specifically ferret, fruit bat and civet ACE2 which appear to be 204

preferentially used by SARS-CoV ( Fig.2A and Sup. Fig.5B ). In summary, using two distinct 205

technical approaches that monitor Spike-mediated receptor usage in a biologically relevant 206

context we provide evidence that SARS-CoV-2 has a broad tropism for mammalian ACE2s. 207

These assays demonstrate correlation between ACE2 protein sequence and fusion by SARS-208

CoV or SARS-CoV-2 Spike protein, plus evidence of a low affinity of SARS-CoV Spike proteins 209

for bird or rat ACE2 and varying levels of bat ACE2 utilisation. 210

High throughput and robust, surrogate assays for SARS-CoV-2 viral entry only serve to model 213 this process and can never completely replace live virus experiments. To this end, and in order 214

to examine the permissiveness of non-human cell lines in our cell culture collection 215 (Sup. We next sought to correlate the receptor usage results from our surrogate entry assays (Fig.2 ) 229

with live virus infections. The hamster kidney cell line BHK-21, which we established as 230 refractory to SARS-CoV-2 infection (Fig.3A,B ), was transfected with vector alone (pDISPLAY) 231

or a restricted panel of ACE2 constructs (hamster, human, horseshoe bat, rabbit, pig and 232 chicken) representing the spectrum of receptor usage ( Fig.2A) . Concurrent to the infections, 233

the expression of ACE2 in equivalently transfected cells was confirmed by western blot, flow 234 cytometry and SARS-CoV-2 pseudotype infections (Sup. Fig.6A -D and Sup. Fig.1 ). Of note, 235

for the live virus infections the high MOI (1) inoculum was removed after 1 hour with the cells 236

thoroughly washed prior to incubation at 37 °C. Accordingly, in the BHK-21 cells transfected 237 with carrier plasmid we saw very little evidence for virus infection and/or virus production, 238

confirming these cells do not natively support SARS-CoV-2 infection (Fig.3C ). For the 239 receptors where we had previously seen high levels of cell-cell fusion (hamster, pig and rabbit) 240

we observed robust viral replication (Fig.3C) . Surprisingly, the two receptors included because 241 of their 'poor' usage by SARS-CoV-2 Spike (horseshoe bat and chicken ACE2, Fig.2A ) were 242 still able to support viral replication, albeit to a lower level. Of note, regardless of the ACE2 243 species expressed we saw very little evidence of cytopathic effect in the infected BHK-21 cells 244

(Sup. Fig.6E ), despite the release of infectious virus into the supernatant (Sup. Fig.6F ). Lastly, 245

focusing on the unexpected observation that chicken ACE2 permitted SARS-CoV-2 entry into 246 cells, we investigated whether chicken DF-1 cells over-expressing ACE2 could support viral 247

replication. Whilst western blot and flow cytometry demonstrated successful ACE2 over-248 expression (Sup. Fig.6B ,D) we did not see any evidence of viral replication in these cells, 249

either because of inefficient chicken ACE2 receptor usage or a post-entry block to SARS-CoV-250 2 replication (Fig.3D ). In summary, SARS-CoV-2 is able to use a range of non-human ACE2 251 receptors to enter cells. Furthermore, when a cognate ACE2 is provided the virus can replicate 252 efficiently in the normally refractory hamster cell line BHK-21.

Discussion 255 256

Recognising animals at risk of infection and/or identifying the original or intermediate hosts 257

responsible for the SARS-CoV-2 pandemic are important goals for ongoing research. In addition, there is a requirement to develop appropriate animal models for infection 259 that, if possible, recapitulate the hallmarks of disease seen in people. Importantly, high-260 resolution structures of human ACE2 in complex with the Spike RBD [5, 6, 10, 11] can help 261 us to understand the genetic determinants of SARS-CoV-2 host-range and pathogenesis. In 262 particular, differences in receptor usage between closely related species provides an 263 opportunity to pinpoint amino acid substitutions at the interaction interface that inhibit Spike 264 protein binding and thus fusion.

One example of closely related ACE2 sequences differing in their utilisation by SARS-CoV-2 267

Spike comes from the comparison of rat and hamster ACE2. Although a number of animal 268 models have been investigated for SARS-CoV-2, including non-human primates, ferrets and 269 cats [18, 19] , the use of small animals, in particular rodents, has proved more challenging as 270 murine and rat ACE2 support lower levels of β-coronavirus entry [1, 15] . For SARS-CoV this 271

problem was circumvented with the development of transgenic mice expressing human ACE2 272

[20] or mouse-adapted SARS-CoV [21, 22] . Consistent with previously published data on 273 SARS-CoV rodent ACE2 interactions, we showed that rat ACE2 does not support SARS-CoV-274 2 mediated fusion ( Fig.2A) . However, our finding that hamster ACE2 allows entry of SARS-275

CoV-2 ( Fig interface that might explain this variable receptor tropism (listed as hamster to rat): Q24K, 279

T27S, D30N, L79I, Y83F, K353H. Except for L79I, which is similarly substituted in pangolin 280

and pig ACE2, all of these substitutions are likely to reduce Spike RBD binding. Q24K and 281

Y83F substitutions would both result in the loss of hydrogen bonds with the side chain of 282 SARS-CoV-2 RBD residue N487 (Fig.4B ). Residue D30 is acidic in all ACE2 proteins that are 283 efficiently utilised by SARS-CoV-2 Spike, and its substitution to asparagine would remove the 284 salt bridge formed with K417 of the RBD. Lastly, the T27S substitution would remove the 285 threonine side chain methyl group that sits in a hydrophobic pocket formed by the side chains 286

of RBD residues F456, Y473, A475 and Y489. Thus, multiple substitutions are predicted to 287

inhibit Spike binding to rat ACE2 when compared with the closely related hamster protein. Of 288 note, the hamster cell lines we used in our study (BHK-21 and CHO) are likely refractory to 289 infection simply because they express low levels of ACE2 mRNA (Fig.3A , qPCR data). 290

Interestingly, the high level of cell-cell fusion seen with rabbit ACE2 indicates that lagomorphs 291 may also represent a good model organism for SARS-CoV-2 pathogenesis. 292 293

A second example of different receptor usage between closely related species can be seen 294

with bat ACE2 ( Fig.2A, 4A) . The apparent lack of tropism for bat ACE2 proteins we observed 295

was surprising as there is previous evidence of SARS-CoV-2 infection of bat ACE2 expressing 296 cells in vitro [1] and in vitro binding experiments suggest that the SARS-CoV-2 RBD binds bat 297

ACE2 with high affinity [24] . Since the exact origin of SARS-CoV-2 is currently unknown, but 298

widely accepted to be a Chiroptera species, we included ACE2 proteins from a broad range 299 of bats in our study. While none support SARS-CoV-2 fusion to the same levels as humans, 300

there are dramatic differences in the ability of SARS-CoV-2 Spike to utilise ACE2 from 301 horseshoe bats versus fruit bats and little brown bats ( Fig.2A ). As discussed earlier, the 302 closest known relative of SARS-CoV-2, RaTG13, was isolated from intermediate horseshoe 303

bat (Rhinolophus affinis). Unfortunately, the ACE2 sequence from this species was not 304 available for use in our study; however, we did include an ACE2 from the closely related least 305

horseshoe bat (Rhinolophus pusillus). Although this protein supported the lowest levels of 306

fusion of any bat ACE2 tested in our study, it still supported a low level of SARS-CoV-2 307 replication with live virus (Fig.3C ). As in rat ACE2, horseshoe bat and fruit bat ACE2 have a 308 lysine residue at position 24 that would disrupt hydrogen bonding to N487 of the SARS-CoV-309

2 RBD and introduce a charge (Fig.4A ,B). Little brown bats have the hydrophobic residue 310 leucine at this position, which could not form the hydrogen bond to N487 but which is present 311

in ACE2 from several species that support high levels of fusion, suggesting that loss of the 312 hydrogen bond is less deleterious to Spike protein binding than introduction of the lysine 313 positive charge. Fruit bats conserve a T27 whereas little brown bats have the bulkier isoleucine 314 residue and horseshoe bats have a bulky charged lysine residue in this position, both of which 315 are likely to clash with the F456-Y473-A475-Y489 hydrophobic pocket of the RBD, with the 316 lysine substitution likely to be more deleterious due to the introduction of the positive charge. 317

Like rats, horseshoe bat N30 would be unable to form a salt bridge with RBD K417. 318

Substitution of Q42 with glutamate in little brown bat may be detrimental to Spike binding as 319

it would disrupt the hydrogen bond to the backbone carbonyl oxygen of RBD residue G446. 320

The other substitutions between bat ACE2 proteins and other mammals are likely to be benign. 321

Little brown bats, horseshoe bats, pangolins and horses all share a serine as ACE2 residue 322 34, suggesting that serine in this position does not abolish Spike binding, and it is likely that 323 the threonine at this position (fruit bat ACE2) would likewise be tolerated. Similarly, the Y41H 324 substitution present in little brown bat ACE2 is also present in horse ACE2, suggesting that it 325

does not prevent binding. Therefore, all bat ACE2 proteins have substitutions that impair 326 SARS-CoV-2 Spike binding to different degrees, but it seems likely that the E30N substitution 327 (shared only by rat ACE2) is the most likely cause of the severely impaired binding of SARS-328

CoV-2 Spike to horseshoe bat ACE2.

Interestingly, a similarly 'poor' tropism for bat ACE2 was also reported for SARS-CoV following 331

its be possible to easily identify the animal or animals that seeded the original epidemic.

A third example of ACE2 usage by SARS-CoV-2 differing dramatically between closely related 356 species is dog and ferret. It was surprising that entry of SARS-CoV-2 pseudotypes into cells 357 was heavily restricted by ferret ACE2 (c. 1% of human levels, Fig.2A charged residue arginine in ferret is likely to decrease binding efficiency, although we note 361

that pangolin ACE2 has a His residue in this position and retains binding to Spike (Fig.4A,B) .

Similarly, substitution of L79 in dog ACE2 with histidine in ferret would be likely to disrupt 363 hydrophobic interactions with the side chain of Spike F486. Comparison of mammalian ACE2 364 receptors usage by SARS-CoV versus SARS-CoV-2 can also be coupled to inspection of the 365 available ACE2:RBD co-structures [5, 6, 10, 11, 28 ] to obtain further molecular insights into 366

Spike binding. The long arginine side chain of SARS-CoV residue 426, which is the structural 367 equivalent to SARS-CoV-2 N439, makes a salt bridge with E329 of human ACE2 and interacts 368

with the side chain of ACE2 Gln325. This results in a larger binding footprint for SARS-CoV 369 on human ACE2 when compared to SARS-CoV-2 (see, for example, Figure 3 in [6]). It is 370 therefore striking that ferret ACE2 is not used efficiently by SARS-CoV-2 for fusion, while ferret 371 ACE2 can support SARS-CoV-mediated fusion. The enhanced usage may arise from a salt 372 bridge being formed between ferret ACE2 residue E325 and SARS-CoV R426, which is not 373 possible in SARS-CoV-2 where the equivalent residue is asparagine. This additional salt 374 bridge may therefore 'rescue' some of the binding loss caused by the deleterious substitutions While not as dramatic as for the species listed above, differences in SARS-CoV-2 receptor 384 utilisation are also seen for cat versus civet ACE2. The most likely candidate causative 385 substitution in civet ACE2 is K31T, where a complementary long-range charge interaction with 386 RBD E484 is lost. In SARS-CoV the entire loop between residues 460-472 (equivalent to 387 SARS-CoV-2 residues 473-486) is reordered and there is not a glutamate or aspartate residue 388 in this local vicinity. We would therefore not predict cat ACE2 to bind SARS-CoV Spike more 389 strongly than civet ACE2, consistent with our fusion assay data. 390 391

In the process of finalising this manuscript two papers were released as preprints, also 392 examining the receptor usage of various non-human ACE2s with surrogate virus entry assays 393

(lentiviral pseudotypes) [29, 30] . While these studies did not perform corresponding 394

examination of cell-cell fusion or follow up experiments with SARS-CoV-2 live virus there is a 395 strong correlation between their findings and ours, namely the broad tropism of SARS-CoV-2 396

Spike. Notably, all three research data sets concur that human and several non-human ACE2 397

proteins support similar levels of utilisation by SARS-CoV-2 Spike, in contrast to a recent 398

report that claimed preferential binding to the human ACE2 VLPs may more accurately reflect the number and conformation of viral proteins in the live 411 virus particle they cannot easily be manipulated to encode a reporter gene, such as Firefly 412

luciferase. As such our cell-cell system or indeed live virus may therefore be more appropriate 413

for probing low affinity interactions between atypical host ACE2s and coronavirus Spike 414

proteins. While more evidence is required to examine the significance of cell-cell spread 415

(syncytia formation) of SARS-CoV-2 in vitro and in vivo, the quantitative assay we have 416 developed promises to be a robust tool for supporting these efforts. Similarly, examining 417

whether the polybasic cleavage site found in SARS-CoV-2 Spike provides a selective 418 advantage to this virus, e.g. by allowing enhanced spread of the virus through cell-cell fusion, 419

is the source of ongoing investigations in our laboratory. the epidemiological significance of these infections remains to be determined, for example 429 whether they represent one-off spill-over events without onward transmission, or alternatively 430 evidence for the existence of animal reservoirs. The latter scenario would have significant 431 epidemiological implications to human populations recovering from the first wave of the SARS-432

CoV-2 pandemic. Interestingly, certain animals where we demonstrated efficient receptor 433

usage, e.g. pigs and dogs ( Fig.2A) Cell lines 449

Cell lines representing a broad range of animal species were used to determine the host 450 range/tropism of SARS-CoV-2 (Sup. cells were infected at MOI 1 as described above and supernatants collected at 72h post 498

infection and frozen at -80°C until required. 499 500

Pseudoparticle generation: Lentiviral based pseudoparticles were generated in HEK293T 502 producer cells, seeded in 6-well plates at 7.5x10 5 /well one day prior to transfecting with the 503 following plasmids: 600ng p8.91 (encoding for HIV-1 gag-pol), 600ng CSFLW (lentivirus 504 backbone expressing a firefly luciferase reporter gene) and either 25ng of SARS-CoV-2 Spike 505 or 500ng SARS-CoV in OptiMEM (Gibco) (Sup. Table. 3) with 10µL PEI, 1µg/mL (Sigma) 506 transfection reagent. No glycoprotein controls (NE) were also set up using empty plasmid 507 vectors (25ng pCAGGS for SARS-CoV-2 and 500ng pcDNA3.1 for SARS-CoV) and all 508 transfected cells were incubated at 37°C, 5% CO2. The following day, the transfection mix was 509

replaced with DMEM-10% and pooled harvests of supernatants containing SARS-CoV-2 510 pseudoparticles (SARS-CoV-2 pps) and SARS-CoV pseudoparticles (SARS-CoV pps) were 511 taken at 48 and 72h post transfection, centrifuged at 1,300 x g for 10 mins at 4°C to remove 512 cellular debris, aliquoted and stored at -80°C. HEK293T target cells transfected with 500ng of 513 a human ACE2 expression plasmid (Addgene) were seeded at 2x10 4 in 100μL DMEM-10% in 514 a white-bottomed 96-well plate (Corning) one day prior to infection. SARS-CoV-2 pp and 515 SARS-CoV pp along with their respective NE controls were titrated 10-fold on target cells and 516

incubated for 72h at 37°C, 5% CO2. To quantify firefly luciferase, media was replaced with 517

50µL Bright-Glo™ substrate (Promega) diluted 1:2 with serum-free, phenol red-free DMEM, 518 incubated in the dark for 2 mins and read on a Glomax Multi+ Detection System (Promega). 519 520

Receptor usage screens: BHK-21 cells were seeded in 48-well plates at 5x10 4 /well in DMEM-521 10% one day prior to transfection with 500ng of different species ACE2-expression constructs 522 or empty vector (pDISPLAY) (Sup. Table. 2) in OptiMEM and TransIT-X2 (Mirus) transfection 523 reagent according to the manufacturer's recommendations. The next day, cells were infected 524 with SARS-CoV-2 pp/SARS-CoV pp equivalent to 10 6 -10 7 relative light units (RLU), or their 525 respective NE controls at the same dilution and incubated for 48h at 37°C, 5% CO2. To 526 quantify Firefly luciferase, media was replaced with 100µL Bright-Glo™ substrate (Promega) 527 diluted 1:2 with serum-free, phenol red-free DMEM. Cells were resuspended in the substrate 528

and 50µL transferred to a white-bottomed plate in duplicate. The plate was incubated in the 529 dark for 2 mins then read on a Glomax Multi+ Detection System (Promega) as above. CSV 530 files were exported onto a USB flash drive for analysis. Biological replicates were performed 531 three times. 532 533

Cell-cell fusion assays 534

HEK293T rLuc-GFP 1-7 [40] effector cells were transfected in OptiMEM (Gibco) using Transit-535

X2 transfection reagent (Mirus), as per the manufacturer's recommendations, with SARS-536

CoV-2 (250ng), SARS-CoV (1000ng) (Sup. Table. 3) or mock-transfected with empty plasmid 537 vector (pCAGGS for SARS-CoV-2 and pcDNA3.1+ for SARS-CoV). BHK-21 target cells were 538 co-transfected with 500ng of different ACE2-expressing constructs (Sup. Table. 2) and 250ng 539

of rLuc-GFP 8-11 plasmid. For SARS-CoV-2 cell-cell fusion assays, target cells were also 540 transfected with 25ng of transmembrane protease serine 2 (TMPRSS2) for 48h. SARS-CoV-541 2 effector cells were washed once with PBS and resuspended in phenol red-free DMEM-10%.

SARS-CoV effector cells were washed twice with PBS and incubated with 3µg/ml of TPCK-543 treated trypsin (Sigma-Aldrich) for 30 mins at 37°C before resuspending in phenol red-free 544 DMEM-10%. Target cells were washed once with PBS and harvested with 2mM EDTA in PBS 545 before co-culture with effector cells at a ratio of 1:1 in white 96-well plates to a final density of 546 4×10 4 cells/well in phenol red-free DMEM-10%. Quantification of cell-cell fusion was measured 547

based on Renilla luciferase activity, 18h (SARS-CoV-2) or 24h (SARS-CoV) later by adding 548

1µM of Coelenterazine-H (Promega) at 1:400 dilution in PBS. The plate was incubated in the 549 dark for 2 mins then read on a Glomax Multi+ Detection System (Promega) as above. CSV 550 files were exported onto a USB flash drive for analysis. GFP fluorescence images were 551 captured every 2h for 24h using an Incucyte S3 real-time imager (Essen Bioscience, Ann 552

Arbor, MI, USA). Cells were maintained under cell culture conditions as described above. 553

Assays were set up with three or more biological replicates for each condition, with each 554 experiment performed three times.

Western blotting 557 BHK-21 cells were transfected using Transit-X2 transfection reagent (Mirus), as per the 558 manufacturer's instructions with 500ng of different ACE2-expression constructs (Sup. Table. 2) 559

or mock-transfected with empty plasmid vector (pDISPLAY). All protein samples were 560 generated using 2x Laemmli buffer (Bio-Rad) and reduced at 95°C for 5 mins 48h post-561

transfection. Samples were resolved on 7.5% acrylamide gels by SDS-PAGE, using semi-dry 562 transfer onto nitrocellulose membrane. Blots were probed with mouse anti-HA primary 563 antibody (Miltenyi Biotech) at 1:1,000 in PBS-Tween 20 (PBS-T, 0.1%) with 5% (w/v) milk 564 powder overnight at 4°C. Blots were washed in PBS-T and incubated with anti-mouse 565 secondary antibody conjugated with horseradish peroxidase (Cell Signalling) at 1:1,000 in 566 PBS-T for 1h at room temperature. Membranes were exposed to Clarity Western ECL 567 substrate (Bio-Rad Laboratories) according to the manufacturer's guidelines and exposed to 568 autoradiographic film.

Flow cytometry 571 BHK-21 cells were transfected using Transit-X2 transfection reagent (Mirus), as per the 572 manufacturer's instructions with 500ng of each ACE2-expression construct (Sup. Table. 2) or 573 mock-transfected with empty plasmid vector (pDISPLAY) for 48h. Cells were resuspended in 574 cold PBS and washed in cold stain buffer (PBS with 1% BSA (Sigma-Aldrich), 0.01% NaN3 575

and protease inhibitors (Thermo Scientific)). Cells were stained with anti-HA PE-conjugated 576 antibody (Miltenyi Biotech) at 1:50 dilution for 1×10 6 cells for 30 mins on ice, washed twice 577

with stain buffer and fixed in 2% paraformaldehyde for 20 mins on ice. Fixed cells were 578 resuspended in PBS before being analysed using the MACSQuant® Analyzer 10 (Miltenyi 579

Biotech) and the percentage of PE-positive cells was calculated by comparison with unstained 580

and stained mock-transfected samples. Positive cells were gated as represented in Sup. Fig.1  581 and the same gating strategy was applied in all experiments. 582 583 RNA extraction and ACE2 qPCR quantification 584

Total cellular RNA was extracted from cell lines in Sup. Table. 1 using a QIAGEN RNeasy RNA 585 extraction kit and mRNA was then detected with SYBR-green based qPCR, using a standard 586 curve for quantification on a Quant studio 3 thermocycler. Luna® Universal qPCR Master Mix 587

(NEB) was used to quantify mRNA levels for each cell line. RNA was first transcribed using 588

SuperScript II Reverse Transcriptase (Thermo Fisher), with oligo dT primers and 50ng of input 589 RNA in each reaction. All the reactions were carried out following the manufacturer's 590

instructions and in technical duplicate, with the melt curves analysed for quality control 591

purposes. Conserved cross-species ACE2 primers used for each cell line are found in 592

Sup. ACE2 amino acid sequences were translated from predicted mRNA sequences or protein 606 sequences (Sup. Table. 2). The predicted guinea pig mRNA sequence was more divergent 607 than expected and contained a premature stop codon. For the purposes of this research, five 608 single nucleotides were added, based on the most closely related sequence (chinchilla), to 609 allow a full-length mature protein to be synthesised. It is not clear if the guinea pig has a 610 functional ACE2, or if the quality of the genomic data is very low, but overall confidence in this 611

sequence is low. The other divergent sequence was turkey as the 3' end was not homologous 612

with other vertebrate ACE2 receptors. This appeared to be a mis-annotation in the genome 613 as the 3' end showed very high identity to the collectrin gene. The missing 3' of the gene was 614

found in the raw genome data and assembled with the 5' region to make a full ACE2 sequence. 615

Twenty-three nucleotide base pairs were missing between these regions; these were taken 616 from chicken as the most closely related sequence. infections were performed in duplicate, with error bars denoting standard deviation from the 708 mean. 709 710 between human ACE2 (green) and SARS-CoV-2 RBD (yellow). Insets 1 to 7 show molecular 716

interactions discussed in the main text. Bonds that may be disrupted are shown as grey lines, 717

with bond distances in grey text, and hydrophobic interactions that may be disrupted are 718 

D.

Pseudotype and cell-cell fusion assays were established for SARS-CoV-2 (A,B) and SARS-CoV (C,D) using multiple internal controls. For the pseudotype assays non-enveloped (NE) lentiviral particles were generated, i.e. vector plasmid in place of a viral glycoprotein, to examine background levels of pseudoparticle entry. For the cell-cell fusion assay mock-transfected effector cells were used (No Spike) to examine background levels of cell-cell fusion. In all subsequent experiments 'NE' and 'No Spike' controls were compared against SARS-CoV-2 pseudoparticles or SARS-CoV-2 Spike expressing effector cells (see Sup. Fig.3) . To validate our pDISPLAY approach cells were transfected with expression constructs for full length human ACE2 (hACE2 [FL]) or a human ACE2 where the signal peptide was replaced with the murine Ig κ-chain leader sequence (hACE2). In both instances the corresponding vector controls, pcDNA3.1 and pDISPLAY, were seperately transfected for comparison. The specificity of the SARS-CoV-2 and SARS-CoV assays were further confirmed by comparing hACE2-mediated fusion to human aminopeptidase N (hAPN) or dipeptidyl peptidase 4 (hDPP4) fusion, the coronavirus group I and MERS-CoV receptors, respectively. Lastly, in all assays target cells representing un-transfected cells (Mock) were also included. For pseudotype and cellcell fusion assays, luciferase assays were performed in duplicate and triplicate, respectively with the error bars denoting standard deviation. Figure 3 : Syncytia formation following SARS-CoV-2 Spike expression. Effector cells expressing half of a split luciferase-GFP reporter and SARS-CoV-2 Spike were mixed with target cells expressing ACE2 proteins from the indicated hosts and the corresponding half of the reporter (see Methods). A vector only control was also included (pDISPLAY). Representative micrographs of GFP-positive syncytia formed following co-culturing are shown. Images were captured using an Incucyte live cell imager (Sartorius). 

A 727 pneumonia outbreak associated with a new coronavirus of probable bat origin

Bats are natural 731 reservoirs of SARS-like coronaviruses

Evidence for camel-to-human transmission of MERS coronavirus

A systematic review of antibody mediated immunity to 738 coronaviruses: antibody kinetics, correlates of protection, and association of antibody 739 responses with severity of disease. medRxiv

Structural 741 basis of receptor recognition by SARS-CoV-2

Structural and Functional Basis of SARS-CoV-2 Entry by Using Human 744 ACE2

Receptor recognition mechanisms of coronaviruses: a decade of structural studies

CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease 750 Inhibitor

Functional assessment of cell entry and receptor usage 752 for SARS-CoV-2 and other lineage B betacoronaviruses

754 Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 receptor. 755 Nature

Structural basis for the recognition of SARS-757

CoV-2 by full-length human ACE2

Mechanisms of host receptor adaptation 759 by severe acute respiratory syndrome coronavirus

Predicting the angiotensin 761 converting enzyme 2 (ACE2) utilizing capability as the receptor of SARS-CoV-2. Microbes Infect

Structure-Guided Identification of a Nonhuman Morbillivirus with Zoonotic Potential

Efficient replication of severe acute respiratory syndrome coronavirus in mouse 769 cells is limited by murine angiotensin-converting enzyme 2

Structure, Function, 771 and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Enhanced isolation of SARS-CoV-2 by TMPRSS2-expressing cells

Infection and Rapid Transmission 779 of SARS-CoV-2 in Ferrets

Susceptibility 782 of ferrets, cats, dogs, and other domesticated animals to SARS-coronavirus 2

Severe acute respiratory 785 syndrome coronavirus infection causes neuronal death in the absence of encephalitis in mice 786 transgenic for human ACE2

788 A new mouse-adapted strain of SARS-CoV as a lethal model for evaluating antiviral agents in 789 vitro and in vivo

A mouse-adapted SARS-coronavirus 792 causes disease and mortality in BALB/c mice

CoV-2 in golden hamsters

Characterization of the 797 receptor-binding domain (RBD) of 2019 novel coronavirus: implication for development of RBD 798 protein as a viral attachment inhibitor and vaccine

Difference in receptor usage between severe acute respiratory syndrome (SARS) coronavirus 801 and SARS-like coronavirus of bat origin

Structural analysis of major species barriers between humans and palm civets for severe 803 acute respiratory syndrome coronavirus infections

Probable Pangolin Origin of SARS-CoV-2 Associated with the 805 COVID-19 Outbreak

Structure of SARS coronavirus spike receptor-binding 807 domain complexed with receptor

Potential host range of multiple SARS-like coronaviruses and an improved ACE2-Fc 810 variant that is potent against both SARS-CoV-2 and SARS-CoV-1. bioRxiv

Broad and differential animal ACE2 receptor usage by SARS-CoV-2. 813 bioRxiv

In silico comparison of spike protein-815 ACE2 binding affinities across species; significance for the possible origin of the SARS-CoV-2 816 virus

Neuropilin-1 is 820 a host factor for SARS-CoV-2 infection. bioRxiv

Activation of the SARS coronavirus spike protein 822 via sequential proteolytic cleavage at two distinct sites

The S2 Subunit of 825 Infectious Bronchitis Virus Beaudette Is a Determinant of Cellular Tropism

Fusion-defective mutants of mouse hepatitis virus 827 A59 contain a mutation in the spike protein cleavage signal

The S2 subunit of the spike glycoprotein of bovine 829 coronavirus mediates membrane fusion in insect cells

Canine SARS-CoV-2 infection. PREPRINT (Version 1) 833 available at Research Square

SARS-CoV-2 neutralizing serum antibodies in cats: a 836 serological investigation. bioRxiv, 2020. 837 39

Nectin-4, a novel peste des petits ruminants virus receptor

Generation of a dual-functional 840 split-reporter protein for monitoring membrane fusion using self-associating split GFP

a new generation of protein database search programs. Nucleic 844 Acids Res

UniRef clusters: a comprehensive 846 and scalable alternative for improving sequence similarity searches

MAFFT: a novel method for rapid multiple 849 sequence alignment based on fast Fourier transform

852 an improved methodology to estimate and visualize evolutionary conservation in 853 macromolecules

The neighbor-joining method: a new method for reconstructing 855 phylogenetic trees

MEGA7: Molecular Evolutionary Genetics Analysis 857 Version 7.0 for Bigger Datasets

WebLogo: a sequence logo generator

BHK-21 cells were transfected with a panel of species-specific ACE2-expressing constructs (see Sup.Table.2). Cells were surface stained with anti-HA PE conjugated antibody. Live and singlet BHK-21 were gated as PE-positive, relative to mock

Representative datasets are shown for human, goat and guinea pig ACE2 surface staining (bottom panel)