key: cord-1007015-yqeifpoy
authors: Zhai, Xiaofeng; Sun, Jiumeng; Yan, Ziqing; Zhang, Jie; Zhao, Jin; Zhao, Zongzheng; Gao, Qi; He, Wan-Ting; Veit, Michael; Su, Shuo
title: Comparison of SARS-CoV-2 spike protein binding to human, pet, farm animals, and putative intermediate hosts ACE2 and ACE2 receptors
date: 2020-05-08
journal: bioRxiv
DOI: 10.1101/2020.05.08.084061
sha: e7210646c4d032b5150779d69df9abee452eb49c
doc_id: 1007015
cord_uid: yqeifpoy

The emergence of a novel coronavirus, SARS-CoV-2, resulted in a pandemic. Here, we used recently released X-ray structures of human ACE2 bound to the receptor-binding domain (RBD) of the spike protein (S) from SARS-CoV-2 to predict its binding to ACE2 proteins from different animals, including pets, farm animals, and putative intermediate hosts of SARS-CoV-2. Comparing the interaction sites of ACE2 proteins known to serve or not serve as receptor allows to define residues important for binding. From the 20 amino acids in ACE2 that contact S up to seven can be replaced and ACE2 can still function as the SARS-CoV-2 receptor. These variable amino acids are clustered at certain positions, mostly at the periphery of the binding site, while changes of the invariable residues prevent S-binding or infection of the respective animal. Some ACE2 proteins even tolerate the loss or the acquisition of N-glycosylation sites located near the S-interface. Of note, pigs and dogs which are not or not effectively infected, respectively, have only a few changes in the binding site have relatively low levels of ACE2 in the respiratory tract. Comparison of the RBD of S of SARS-CoV-2 with viruses from bat and pangolin revealed that the latter contains only one substitution, whereas the bat virus exhibits five. However, ACE2 of pangolin exhibit seven changes relative to human ACE2, a similar number of substitutions is present in ACE2 of bats, raccoon, and civet suggesting that SARS-CoV-2 may not especially adapted to ACE2 of any of its putative intermediate hosts. These analyses provide new insight into the receptor usage and animal source/origin of SARS-COV-2. IMPORTANCE SARS-CoV-2 is threatening people worldwide and there are no drugs or vaccines available to mitigate its spread. The origin of the virus is still unclear and whether pets and livestock can be infected and transmit SARS-CoV-2 are important and unknown scientific questions. Effective binding to the host receptor ACE2 is the first prerequisite for infection of cells and determines the host range. Our analysis provides a framework for the prediction of potential hosts of SARS-CoV-2. We found that ACE2 from species known to support SARS-CoV-2 infection tolerate many amino acid changes indicating that the species barrier might be low. However, the lower expression of ACE2 in the upper respiratory tract of some pets and livestock means more research and monitoring should be done to explore the animal source of infection and the risk of potential cross-species transmission. Finally, the analysis also showed that SARS-CoV-2 may not specifically adapted to any of its putative intermediate hosts.

infection of the respective animal. Some ACE2 proteins even tolerate the loss or the acquisition of N-23 glycosylation sites located near the S-interface. Of note, pigs and dogs which are not or not effectively 24 infected, respectively, have only a few changes in the binding site have relatively low levels of ACE2 in 25 the respiratory tract. Comparison of the RBD of S of SARS-CoV-2 with viruses from bat and pangolin 26 revealed that the latter contains only one substitution, whereas the bat virus exhibits five. However, 27 ACE2 of pangolin exhibit seven changes relative to human ACE2, a similar number of substitutions is 28 present in ACE2 of bats, raccoon, and civet suggesting that SARS-CoV-2 may not especially adapted to 29 ACE2 of any of its putative intermediate hosts. These analyses provide new insight into the receptor 30 usage and animal source/origin of SARS-COV-2. 31 IMPORTANCE 32 SARS-CoV-2 is threatening people worldwide and there are no drugs or vaccines available to mitigate 33 its spread. The origin of the virus is still unclear and whether pets and livestock can be infected and 34 transmit SARS-CoV-2 are important and unknown scientific questions. Effective binding to the host 35 receptor ACE2 is the first prerequisite for infection of cells and determines the host range. Our analysis 36 provides a framework for the prediction of potential hosts of SARS-CoV-2. We found that ACE2 from 37 species known to support SARS-CoV-2 infection tolerate many amino acid changes indicating that the 38 species barrier might be low. However, the lower expression of ACE2 in the upper respiratory tract of

As of the 30 th of April 2020, the ongoing pandemic of a novel coronavirus, severe acute respiratory 45 syndrome coronavirus 2 (SARS-CoV-2), has developed into a global challenge with the number of total 46 confirmed cases exceeding 3 million including more than 200 thousands fatalities, thereby causing a 47 major loss to global public health and the world economy. This disease is referred to as the 2019 48 coronavirus disease by World Health Organization (WHO) and was defined as a public 49 health emergency of international concern (PHEIC) on the 30 th of January 2020. Its main clinical 50 symptoms include fever, fatigue, and dry cough. A rather large proportion of patients become critically 51 ill with acute respiratory distress syndrome, similar to patients with severe acute respiratory syndrome 52 (SARS) caused by SARS coronavirus (SARS-CoV) (1-3). SARS emerged in China in [2002] [2003] and also 53 rapidly spread worldwide but was contained by public health measures. It is thought that bats and 54 palm civets are the natural and intermediate reservoirs of 5) . Likewise, research suggests 55 that SARS-CoV-2 might have originated also from bats and that pangolins might be the potential 56 intermediate host (6) (7) (8) (9) . Specifically, SARS-CoV-2 has a high nucleotide sequence identity with Bat-CoV-57 RaTG13-like virus except for the middle part of its genome encoding the spike protein which might 58 have derived via recombination from a Pangolin-CoV-like virus (6, 7, (10) (11) (12) . A previous study showed 59 that SARS-CoV-2 replicates poorly in dogs, pigs, but cats are permissive to infection (13). However, 60 whether pets can become new hosts of SARS-CoV-2 needs to be clarified further. 61

The structure of the trimeric spike protein (S) of SARS-CoV-2, the major factor that determines cell and 62

It has been shown that transfection of HeLa cells with genes encoding the human, pig, civet, and bat 118 ACE2 receptor makes them susceptible to infection with SARS-CoV-2, but not with ACE2 from mice 119 (12). To estimate which of the interacting amino acids in ACE2 are essential for binding of S, we 120 compared the ACE2 sequences from humans with that from the other species. Table 1 shows the 121 amino acids that make contacts with S from both SARS-CoV-2 and SARS-CoV (red numbers), only with 122 S from SARS-CoV (blue numbers), only with S from SARS-CoV-2 (green numbers) as well as some 123 variable amino acids in the vicinity (black numbers), and some encoding N-glycosylation sites 124 (highlighted in grey). 125

Pig ACE2 contains five amino acid substitutions in the interacting surface with S from SARS-CoV-2 126 relative to human ACE2 ( Fig. 2A) . Three of the exchanges are located at the periphery of the binding 127 site. Leu79 and Met82, which interact with Phe486 in S, are conservatively substituted by Ile and Thr, 128 respectively. Gln24, which forms a hydrogen bond to N487 in S, is replaced by Leu, which cannot form 129 hydrogen bonds. His34 in the centre of the binding site is also substituted by a Leu residue, which is 130 larger, but cannot from a hydrogen bond. Furthermore, Asp30, which forms the central salt bridge, is 131 exchanged by a Glu residue. Since Glu is also negatively charged, the salt bridge to Lys417 most likely 132 remains intact. It might even become stronger since the side chain of Glu is larger and hence the 133 distance between negatively and positively charged residues becomes smaller. In addition, the N-134 glycosylation site Asn90 in human ACE2 is destroyed by an exchange to Thr. 135

Three amino acid sequences of the ACE2 gene from the bat species Rhinolophus sinicus (R. sinicus) are 136 present in the database. The bats were sampled from R. sinicus colonies in three Chinese provinces 137 including, Guangxi (R. S. -GX), Hubei (R. S. -HB), and Yunnan (R. S. -YN) (28, 29). Strikingly, although 138 their overall amino acid identity is very high (99%), they exhibit large amino acid differences in the N-139 terminal amino acids that contact the S protein (Table 1) . Since the accession number of the bat ACE2 140 is not specified in the publication that demonstrates that it confers susceptibility to SARS-CoV-2 141 infection, it is presently unclear which of them is recognized by the viral spike protein (12). In any case, 142 bat ACE2 proteins contain at least five changes relative to human ACE2. Three of them are substituted 143 in all bat ACE2 sequences and involve the same residues as in pig ACE2, but they are replaced by other amino acids. Gln24 is replaced by the negatively charged Glu or by a positively charged Arg; His34 is 145 substituted by Thr or Ser; and Met82 is substituted by an Asn. In one bat sequence (R. S. -YN) the next 146 but one Pro residue 84 is also replaced by a Ser, which creates a new N-glycosylation site (NXS) at 147 position 82. In addition, in two bat sequences (R. S. -YN and R. S. -GX) Thr27 is substituted by Met or 148 Iso and Tyr41, which forms a hydrogen bond with Thr500 in S, is replaced by the smaller His residue. 149

The most drastic substitution in these two bat ACE2 sequences is Glu35, which forms a hydrogen bond 150 with Gln493 in S, by a positively charged Lys residue. The ACE2 sequence obtained from a bat in the 151

Hubei province, which exhibits the most amino acid substitutions relative to the two other ones, does 152 not contain the latter three changes, but instead Arg31 is replaced by a negatively charged Glu and 153

Asp38 by Asn. The reason why bats exhibit so many changes in residues that interact with S is striking 154 and requires further investigation. However, it is tempting to speculate that local co-evolution 155 between bats and coronaviruses drive these amino acid changes. 156

To get further insight into the amino acids not essential for binding to S, we analysed the ACE2 from 157 civets (Paguma larvata) that has been shown to serve as receptor for SARS-CoV-2 (12). Civet ACE2 158 contains seven amino acid changes relative to human ACE2 (Fig. 2B ). Three of them (Gln24Leu, 159

Asp30Glu, and Met82Thr) are identical to the substitutions in pig ACE2. Another (His34) is at the same 160 position, but exchanged to a different amino acid, Tyr instead of Leu. Another unique, but conservative 161 substitution, Leu45V, is located at the periphery of the binding site, whereas the other three are in the 162 centre. Asp38Glu is a conservative change, but K31T and E37Q replace a charged by an uncharged 163 amino acid. 164

Mouse ACE2, which does not support infection of cells with SARS-CoV-2, has eight amino acid 165 substitutions in the interacting surface with S of SARS-CoV-2 (Fig. 2C) . Three of the sites, Gln24, His34, 166 and Met82 are also replaced in the ACE2 proteins from the two other species and are thus unlikely to 167 be the decisive elements that prevent binding. Leu79 interacts with Phe486 in S (Fig. 6A ) and is 168 exchanged by a Thr. In contrast to bat ACE2, the substitution at position 82 does not create a N-169 glycosylation site in mouse ACE2 since it is exchanged to Ser. Note also that the two used N-170 glycosylation sites near the interacting surface in human ACE2, Asn322, and Asn90, are lost in mouse ACE2 due to exchanges of the Asn residues. The other four exchanges Asp30Asn, Tyr83Phe, Lys31Asn, 172 and Lys353His are more important for preventing binding to mouse ACE2 as discussed in more detail 173 below. 174

In summary, binding of S of SARS-CoV-2 to ACE2 proteins tolerates a surprisingly large number of 175 amino acid changes in the interaction surface, five in pig ACE2 and seven in civet ACE2. Even the 176 acquisition of an additional N-glycosylation site at position 82 due to two substitutions in bat ACE2 177 (M82N, P84S) does not prevent SARS-CoV-2 to use bat ACE2 as receptor in transfected cells. This is in 178 contrast to binding of S of SARS-CoV to rat ACE2, where a glycan attached to the same position 179 prevents binding (23). 180

Next, we asked whether receptor binding might present a species barrier for infection of pets with 182 SARS-CoV-2. Dog ACE2 contains five amino acid changes in the amino acids in contact with S (Fig. 3A) . 183

Residues 24, 30, 34, and 82 are also replaced in pig ACE2, even to the same amino acid at four of the 184 sites. A unique change in dog ACE2 is Asp38Glu, but since this is a conservative change it is unlikely to 185 affect the binding properties of S. Like in pigs, dog ACE2 lacks the glycosylation site at position 90. 186

Asn90 is replaced by Asp in dogs, but to Thr in pigs. Note that the sequence of dog ACE2 in the database 187 contains a non-conservative exchange at position 329 (Glu to Gly), which contacts the S protein of 188 SARS-CoV. This exchange is not present in the ACE2 from a beagle that we sequenced (accession 189 number MT269670). 190

Cat ACE2 contains only four changes (Gln24Leu, Asp30Glu, Asp38Glu, Met82Thr), which are also 191 present in dog ACE2 (Table 1) March (https://www.biorxiv.org/content/10.1101/2020.03.30.015347v1.full.pdf), and it showed that 195 SARS-CoV-2 replicates poorly in dogs, but efficiently in cats and was transmitted by droplets to naïve 196 cats that ACE2 is linked to the fact (13). This indicated that S binding to ACE2 receptors is only the first 197 step in the virus's invasion, and ACE2 levels in different tissues may play an important role in viral transmission. Therefore, the risk of infection in animals requires continuous monitoring. The only 199 residue in dogs, which is not changed in cat ACE2 is His34, which interacts with Tyr453, Leu455, and 200 Gln493 in the centre of the interaction surface. It is exchanged by the slightly later Tyr residue in dog 201 ACE2, which is still able to interact with the same residues in S (Fig. 3A , inset). The other difference is 202 the loop of the N-glycosylation site at position 90. 203

There is also anecdotal evidence that tigers and lions in the Bronx Zoo of New York City were infected 204 by SARS-CoV-2 (https://www.aphis.usda.gov/aphis/newsroom/news/sa_by_date/sa-2020/ny-zoo-205 covid-19). Therefore, we analysed the ACE2 gene from these wild cats. One amino acid difference was 206 detected in ACE2 from cat and tiger, but the residues contacting S are identical explaining why tigers 207 are also susceptible to SARS-CoV-2 infection. ACE2 from a lion has another conservative change 208 relative to cat ACE2: His34 is substituted by Ser and it exhibits a loss of the N-glycosylation site at 209 position 90, like dog ACE2. 210

Exploring potential therapies for COVID-19 animal models is urgently needed. Recent study showed 212 that ferrets to be highly susceptible to infection with SARS-CoV-2 and even transmit virus to naïve 213 contact animals, but also by droplets, albeit the latter route was less efficient (13, 30). Ferret ACE2 214 exhibits the exact same five changes as dog ACE2, but also a substitution of Leu79 by His. In addition, 215 a drastic change occurs at position 354, where the small glycine residue is replaced by a large and 216 positively charged Arg residue (Fig. 3B ). However, the Arg residue avoids clashing with large amino 217 acids in S (Tyr505) by protruding backwards (Fig. 3B , inset). 218

The Syrian hamster (Mesocricetus auratus) is also shown to be susceptible to experimental infection 219 and transmitted SARS-CoV-2 to close contact animals (31). Its ACE2 protein contains only two amino 220 acid substitutions relative to human ACE2. His34 is exchanged to Gln and Met82 is replaced by Asn. 221

Since residue 84 is also exchanged to a Ser, it creates a N-glycosylation site at position 82. A glycan 222 attached to the same position prevents binding of S of SARS to rat ACE2. 223

Guinea pig (Cavia porcellus) might also serve as an animal model, but also common pets, especially of 224 children. Guinea pig ACE2 contains seven amino acid changes at positions 24, 27, 31, 34, 38, 82, and 225 354 and thus more than ACE2 from ferrets or dogs. The glycosylation site at position 90 is preserved, 226 but the site at position 322 is lost due to an Asn to Pro change (Table 1) . 227

Farm animals are also in close contact with humans and thus represent another risk group that might 229 become infected by SARS-CoV-2. The ACE2 proteins from chicken contain ten amino acid changes 230 compared with human ACE2 and lost the N-glycosylation site at position 90 (Fig. 3C) . Some of the 231 affected positions (Gln24, His34, Leu79, Met82, and Gly 353) are also exchanged in ACE2 proteins that 232 serve as SARS-CoV-2 receptor, albeit often to different residues. Unique to all ACE2 proteins is the 233 change of Gln24 by the negatively charged Glu residue, but the interaction with Gly446 and Tyr449 in 234 S is probably preserved (Fig 3C, inset) . Most of the other exchanges are likely to be more critical for 235 binding to S. Tyr at position 83 is replaced by a Phe, which is not able to form a hydrogen bond with S. 236

Two of the changes reverse the polarity of a charged amino acid: Lys31 is substituted by a negatively 237 charged Glu and Glu35 is exchanged to a positively charged Arg. Finally, Asp30, which forms the salt 238 bridge with S, is replaced by the uncharged residue Ala, which makes the ACE2 proteins from chicken 239 and mice the only ones that are not able to form a salt bridge with S. 240

Duck ACE2 also contains ten amino acid substitutions, nine of them at the same position and eight to 241 the same amino, including all the presumably important ones just discussed. Thus, it seems likely that 242 the lack of susceptibility of chicken and ducks to experimental SARS-CoV-2 is due to the inability of the 243 virus to bind to the avian ACE2 receptor (13). 244

The ACE2 protein of pigs can serve as SARS-CoV-2 receptor although it contains five amino acid changes 245 in amino acids contacting the S protein ( Fig. 2A) . Cattle and sheep contain only two amino acid changes 246 (Asp30Glu and Met82Thr) that are also present in pig ACE2 and even retain the N-glycosylation site at 247 position 90 of human ACE2 that is lost in pig ACE2. It thus seems highly likely that ACE2 proteins of 248 both species can function as SARS-CoV-2 receptor and experimental infection of these animals and 249

surveillance is required to show whether they are susceptible to SARS-CoV-2. Camel is the animal 250 reservoir of the Middle East Respiratory syndrome virus (MERS), which however uses dipeptidyl-251 peptidase 4 as protein receptor. We analysed the ACE2 gene from Camelus bactrianus and found, 252 besides the two changes present in cattle and sheep, another substitution: the positively charged Lys31 253 located at the center of the binding site is exchanged by a negatively charged Glu residue, a change 254 which is also present in guinea pig. 255

In summary, almost all mammalian species known to be susceptible to SARS-CoV-2 infection (cats and 256 ferrets) have mutations in many amino acids in their ACE2 proteins. This suggests that these species, 257 especially those in contact with humans, are at risk of contracting the virus and SARS-CoV-2 might 258 establish itself in one of these animals thereby creating an additional animal reservoir. An exception is 259 apparently the pig, which cannot be infected with SARS-CoV-2, although their ACE2 protein can 260 function as SARS-CoV-2 receptor (12). In addition, dogs do not transmit the virus to naïve animals in 261 close contact. We therefore investigated the level of ACE2 expression in different organs by q-RT-PCR. 262

We found that pigs and dogs have the highest mRNA levels in kidney, but the level is also high in other 263

internal organs, such as heart (pigs and dogs) and duodenum and liver (pigs) (Fig. 4) 

Based on an amino acid comparison of ACE2 proteins from animals that do not serve as SARS-CoV-2 276 receptor (mice) or are not susceptible to SARS-CoV-2 infection (chicken) with ACE2 proteins from 277 animals which are infectable (cats, dogs, ferrets) or encode a ACE2 protein that confers susceptibility 278 to SARS-CoV-2 infection (civet), the amino acids essential for binding of S can be deduced (Fig 5, Table 1 ). Tyr83, which forms hydrogen bonds with Asn487 and Tyr489 in S, is replaced by Phe in mice and 280 duck. Phe has the same size and hydrophobicity but is not able to form hydrogen bonds with its side 281 chain. All other animals exhibit at position 83 a Tyr residue. Note, however, that Tyr489 might be 282 shielded by the acquisition of an N-glycosylation site at position 82 as it occurs in ACE2 of mice and in 283 some bat sequences. The other residues are located in the center of the interaction surface. Lys31, 284 which forms van-der-Waals contacts with Y489 and F456 is exchanged by a neutral Asn reside in mice 285 and even to a negatively charged Glu in chicken. Glu35 forms hydrogen bonds with Gln493 in S. It is 286 replaced by a positively charged Arg in chicken, ducks, and some bat sequences. Lys353, which forms 287 hydrogen bonds with its side chain to Tyr495, Gly496, and Gly502, is replaced by the smaller His residue 288 in mice, which presumably cannot form these hydrogen bonds. ACE2 of all other animals (including 289 chicken) contain a Lys at this position. In support of this hypothesis, a single Lys353Ala mutation was 290

shown to abolish the ACE-S interaction (32). Finally, also important is the centrally located Asp30, 291 which is substituted by Asn in mice and chicken. Asn has the same size as Asp, but is uncharged and 292 thus unable to sustain the salt bridge with Lys417 in S. ACE2 proteins from all other animals either 293 retain Asp or it is substituted by the negatively charged, but slightly larger Glu residue, which is 294 probably also able to from the salt bridge. 295

Note that positions 330, 355, and 357 are conserved through all ACE2 proteins we analyzed here and 296 thus their relevance for binding to S cannot be estimated (Table 1 ). In addition, binding of S to ACE2 297 might not follow a simple "lock-and-key" principle. The mouse-adapted SARS-CoV strain MA15 298 contains a single amino acid exchange in the S protein relative to the Urbani strain; Tyr at position 346 299 is replaced by a His residue (33). Tyr 346 forms hydrogen bonds with Asp38 and Gln42 in human ACE2 300 (Fig. 1B) , but mouse ACE2 also contains Asp and Gln at positions 38 and 42, respectively. His is also 301 able to serve as hydrogen-bond donor or acceptor, but its side chain is shorter, and it is not obvious 302 how this exchange enhances binding to mouse ACE2. 303

It has been reported that SARS-CoV-2 derived from a bat virus, but parts of the S protein exhibit a 305 higher nucleotide similarity to a virus from pangolin. Therefore, we aligned the whole amino acid sequence of S from SARS-CoV-2 (first Wuhan isolate) with sequences from the bat and the pangolin 307 virus ( Supplementary Fig. 1 ). As noted before, S from SARS-CoV-2 contains an insertion of four amino 308 acids (PRRA) that creates a polybasic cleavage site recognized by the ubiquitous protease furin (20, 32). 309

Insertion of amino acids at the S1/S2 junction can occur also in bats, since the novel bat-derived virus 310

RmYN02 contains the insertion of amino acid PAA, which does not create a polybasic cleavage site (34). 311

The following S2 subunit is almost completely conserved among all three viruses, it exhibits nine mostly 312 conservative amino acid substitutions in S of the pangolin virus, and only two in S of the bat virus. The 313 N-terminus of the S protein until residue 400 is also highly similar between S of SARS-CoV-2 and the 314 bat virus. There are only six amino acid exchanges in the bat virus, two of them affecting N-315 glycosylation sites, whereas S from the pangolin virus contains 101 amino acid differences compared 316 with S of SARS-CoV-2 ( Supplementary Fig. 1 ). However, from residue 401 to 518, which contains the 317 receptor binding domain of S, the homology reverses (Table 2) . 318

The bat virus contains 18 amino acid substitutions, five of them involve amino acids that contact 319 human ACE2 (white sticks in Fig. 6A ). Located at the periphery of the interaction surface is Phe486, 320 which interacts with Leu79 in ACE2. It is replaced by a Leu, that is also a large and hydrophobic residue. 321

At the other side of the interaction surface located is Asn501, which forms a hydrogen bond with Tyr41 322 in ACE2 and is replaced by the negatively charged Glu residue. The other three exchanged amino acids 323 are located in the centre of the ACE2 binding site. Gln493, which forms a hydrogen bond with Glu35, 324 is replaced by a Tyr residue. Tyr449, which forms a hydrogen bond with Gln42, is replaced by Phe, 325 which has the same size, but cannot form (or only weak) hydrogen bonds. The most drastic exchange 326 is probably Tyr505, which forms a hydrogen bond with Gln42 and is exchanged by a much smaller His 327 residue. 328

In contrast, S from the pangolin virus contains only three exchanges relative to SARS-CoV-2, and only 329 one of them involves an amino acid that contacts ACE2, the substitution of Lys 417 by an Arg (red stick 330 in Fig. 6A ). Since both are basic amino acids, the important salt bridge is probably preserved. In 331 summary, the RBD from the pangolin coronavirus is much better suited to bind to human ACE2 compared to the bat virus. From this point of view, it seems possible that SARS-CoV-2 might have 333 acquired the RBD from a pangolin coronavirus to achieve bat-to-human transmission. 334

We therefore asked whether ACE2 of pangolin (Manis javanica) contains amino acids in its interaction 336 surface that might closely resemble those of humans. In that case, a precursor of SARS-CoV-2 acquired 337 a RBD from a pangolin virus by recombination which is then already adapted to replicate both in 338 pangolins and in humans. However, pangolin ACE2 contains seven changes compared with human 339 ACE2 (Fig. 6B ). In addition, the N-glycosylation site at position 322 of human ACE2 is lost due to an 340 change of Asn to Lys. Some of the changes, Asp30Glu, His34Ser, Asp38Glu, and Leu79Ile, occur also in 341 ACE2 proteins from animals that can interact with S. The three other variable positions are also 342 exchanged in other animals, but mostly to other residues. Gly354 is replaced by a small His residue and 343 not by the larger Arg, that is present in ferret ACE2. Gln24, which interacts with Asn 487 in S, is replaced 344 by a negatively charged Glu residue. Finally, Met82, that interacts with Phe486 in S, is replaced by the 345 larger Asn residue. Although none of the amino acid changes might prevent binding to S, it 346 nevertheless appears that pangolin ACE2 is not especially equipped to serve as receptor for SARS-CoV-347

Other potential intermediate hosts are civets and raccoon dogs (Nyctereutes procyonoides). ACE2 of 349 civets has been shown to confer susceptibility to SARS-CoV-2 infection in cell culture, although it 350 contains seven changes relative to human ACE2 in the amino acids contacting S (Fig. 2B) . The ACE2 351 protein from racoons contains only five substitutions, Gln24Leu, Asp30Glu, His34Tyr, Asp38Glu, and 352

Met82Thr, which are also present in civet ACE2. It is also identical in these residues to ACE2 from dogs, 353 which are susceptible to SARS-CoV-2 infection but do not spread the virus. 354

In summary, none of ACE2 proteins of any of the discussed intermediate hosts seems to be especially 355 equipped to attach to S of SARS-CoV-2, but the least number of changes occur in ACE2 of racoon dogs. 356

It has previously been shown that a RGD motif (403-405: Arg-Gly-Asp) is present in the receptor-359 binding domain of the spike proteins of all SARS-CoV-2 sequences (35). This sequence mediates 360 attachment of several viruses to integrins, which thus might serve as an additional receptor for SARS-361

CoV-2. Two RGD motifs are present in S of pangolin CoV, one at the same site, another at amino acids 362 246-249, but S of bat coronavirus (RaTG13) contains no RGD motif (Fig. 7A) . 363

Bulky carbohydrates attached to the S protein might mask antibody epitopes, interfere with receptor 364 binding and/or with proteolytical cleavage that is required to prime the protein to execute its 365 membrane fusion activity. S of SARS-CoV-2 contains 22 N-glycosylation sites (NXS/T), nine in S2 and 13 366 in the S1 subunit, which are all glycosylated with almost 100% stoichiometry if the ectodomain of S is 367 expressed in mammalian cells (36). 368 A total of 21 glycosylation sites are conserved among all three viruses. S of SARS-CoV-2 contains a 369 unique glycosylation site in the N-terminal domain at position 74 in a loop of the N-terminal S1 domain 370 (Fig. 7B ). This region (aa 60-80) is dissimilar to the pangolin virus, but identical to S of the bat virus, 371 except the glycosylation site (NGI). Two sites are present only in S of the bat and pangolin coronavirus, 372 but not in SARS-CoV-2. One is also located in the N-terminal domain at position 30, in a region (aa 15-373 49) which is identical to S of SARS-CoV-2 virus, except the glycosylation sequon, which is NSS in S of 374 bat, but NSF in S of SARS-CoV-2. The N-terminal domain has a galectin-like folding and is known in 375 other coronaviruses to bind to carbohydrates on the cell surface, i. e. using them as an attachment 376 factor as the first step of virus entry (17). Most interestingly is a site at position 370 which localizes 377 near the ACE2 receptor binding domain. This residue is located in a region (aa 275-400) of high amino 378 acid identity among all three viruses. Thus, it is tempting to speculate that this site was lost during 379 adaption of SARS-CoV-2 to humans, in order to get better access to the ACE2 receptor. 380

As suggested before, the insertion of amino acids at the cleavage site between the S1/S2 subunits of 381 SARS-CoV 2 creates three potential GalNAc O-glycosylation sites, which are not predicted for S of the 382 bat or pangolin virus (37). Attachment of GalNac to serine and threonine residues is catalysed by up to 383 sugar residues are then elongated by other glycosyltransferases thereby creating long carbohydrate 385 chains. Interestingly, whether a certain O-glycoslyation site is used is cell-type dependent and O-386 glycans attached near furin cleavage site have been shown to affect processing (38, 39). It is thus 387 tempting to speculate that the usage of an O-glycosylation site might determine whether S is cleaved 388 in a certain cell. Since processing by furin is essential for entry of SARS-CoV-2 into cells that lack 389 cathepsin proteases, O-glycosylation might affect cell tropism and hence virulence of SARS-CoV-2 (40). 390 This is somewhat reminiscent of hemagglutinin (HA) of an avian Influenza A virus where the loss of a 391 N-glycosylation sequon near a polybasic cleavage site allowed processing of HA thereby generating a 392 highly virulent strain (41). 393

The S protein of SARS-CoV-2 and bat coronavirus RaTG13 were obtained from GenBank (accession 396 numbers QHD43416.1 and QHR63300.2, respectively). The pangolin coronavirus nucleotide sequence 397 was obtained from GISAID (EPI_ISL_410721) and translated into protein. Sequences were aligned using 398 MEGA7.0. A total of 23 ACE2 protein sequences of 20 mammal including human, dog, cat, guinea pig, 399 hamster, mouse, pig, rabbit, cattle. sheep, ferret, raccoon dog, bat (Rhinolophus sinicus), civet, 400 pangolin, tiger, lion, camel, chicken and duck were download from GenBank. The ACE2 protein 401 sequences of dog (MT269670) and cat (MT269670) were obtained from this study (Table 1 ). All the 402 sequences were aligned using MEGA7.0. 403

The software PyMol was used to create the figures of the Cryo-EM structure of S from SARS-CoV-2 405 (pdb file 6VSB) and from the crystal structure of its receptor-binding domain bound to human ACE2 406 (pdb file 6M0J). The amino acids which are mentioned in this publication to mediate contact between 407 S and ACE2 are shown as sticks, whereas the remainder of the molecules are shown as cartoons. The 408 integrated measuring wizard was used to determine the distance between two atoms, which is shown 409 as dotted lines also indicating the distance in Angström. Exchange of certain amino acids was 410 performed with the mutagenesis tool. Among the different possible rotamers of the mutated amino 411 acid side chain the one was chosen that exhibits no clashes with neighboring amino acids. 412 The three amino acids in ACE2 that are not involved in binding of S from SARS-CoV-2 are shown in cyan 595 sticks. The side chain of D30 that forms a salt bridge with S of SARS-CoV-2 is pointing away from the 596 interacting surface. Inset: Detail of the interaction between R426 that forms a salt bridge with E329 597 and a hydrogen bond with Q325. The figure was created with PyMol from pdb file 2AJF. 598 The relative expression of ACE2 in pig was determined in heart, liver, spleen, lung, kidney, duodenum, 610 trachea, turbinate bone, and stomach. (B)The relative expression of ACE2 in dog was determined in 611 heart, liver, spleen, lung, kidney, trachea, and turbinate bone. The experiments were repeated three 612 times.

Amino acids in ACE2 important for binding to S of SARS-CoV-2. Y83 forms hydrogen bonds to 614 N487 and Y489, which are probably weakened by the conservative change to F. K353 forms hydrogen 615 bonds to G502, G496, and Y495, the latter two are probably destroyed by the change to His and the 616 interacting amino acid changes between human and mouse ACE2. Amino acid changes in human ACE2 617 compared with mouse ACE2 are highlighted in red. P84 in human ACE does not interact with amino 618 acids in S of SARS-CoV-2 but might affect the secondary structure. N90T and N 322H destroy the N-619 glycosylation site in human ACE2. 620 

Early 459 Transmission Dynamics in Wuhan, China, of Novel Coronavirus-Infected Pneumonia. The New 460 England journal of medicine

Clinical Characteristics of 138 Hospitalized Patients With

Coronavirus-Infected Pneumonia in Wuhan, China

A 466 familial cluster of pneumonia associated with the 2019 novel coronavirus indicating person-467 to-person transmission: a study of a family cluster

Cross-host evolution of severe acute 474 respiratory syndrome coronavirus in palm civet and human

Discovery of a rich gene pool of bat SARS-related 478 coronaviruses provides new insights into the origin of SARS coronavirus

Probable Pangolin Origin of SARS-CoV-2 Associated with the 488 COVID-19 Outbreak

Identifying SARS-CoV-2 related coronaviruses 491 in Malayan pangolins

Identification of 2019-nCoV related 494 coronaviruses in Malayan pangolins in southern China

A new 497 coronavirus associated with human respiratory disease in China

A pneumonia outbreak 501 associated with a new coronavirus of probable bat origin

Susceptibility of 504 ferrets, cats, dogs, and other domesticated animals to SARS-coronavirus 2

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

Structural basis for the recognition of the 510 SARS-CoV-2 by full-length human ACE2

Structure, Function, 512 and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Structure, Function, and Evolution of Coronavirus Spike Proteins

Highlighted green: amino acids of S of SARS-CoV-2 in contact with ACE2. Highlighted red: amino acid 645 substitutions at these residues in bat or pangolin coronavirus

Pangolin-CoV) S versus SARS-CoV-2 S: one substitution at salt bridge, two 648 substitutions at other sites. Bat coronavirus (Bat-CoV) S versus SARS-CoV-2 S: five substitutions at 649 contact sites