Furin cleavage sites in the spike proteins of bat and rodent coronaviruses: Implications for virus evolution and zoonotic transfer from rodent species.

Bats and rodents comprise two of the world's largest orders of mammals and the order Chiroptera (bats) has been implicated as a major reservoir of coronaviruses in nature and a source of zoonotic transfer to humans. However, the order Rodentia (rodents) also harbors coronaviruses, with two human coronaviruses (HCoV-OC43 and HCoV-HKU1) considered to have rodent origins. The coronavirus spike protein mediates viral entry and is a major determinant of viral tropism; importantly, the spike protein is activated by host cell proteases at two distinct sites, designated as S1/S2 and S2'. SARS-CoV-2, which is considered to be of bat origin, contains a cleavage site for the protease furin at S1/S2, absent from the rest of the currently known betacoronavirus lineage 2b coronaviruses (Sarbecoviruses). This cleavage site is thought to be critical to its replication and pathogenesis, with a notable link to virus transmission. Here, we examine the spike protein across coronaviruses identified in both bat and rodent species and address the role of furin as an activating protease. Utilizing two publicly available furin prediction algorithms (ProP and PiTou) and based on spike sequences reported in GenBank, we show that the S1/S2 furin cleavage site is typically not present in bat virus spike proteins but is common in rodent-associated sequences, and suggest this may have implications for zoonotic transfer. We provide a phylogenetic history of the Embecoviruses (betacoronavirus lineage 2a), including context for the use of furin as an activating protease for the viral spike protein. From a One Health perspective, continued rodent surveillance should be an important consideration in uncovering novel circulating coronaviruses.

Introduction

The Coronaviridae are single-stranded, positive sense, enveloped RNA viruses [1]. They are classified, by genera, as Alpha-, Beta-, Gamma-, and Deltacoronaviruses [2]. The Alpha- and Betacoronaviruses are most often associated with mammalian species, while the Gamma- and Deltacoronaviruses remain largely avian-associated [3]. Across bat and rodent species, both Alpha- and Betacoronaviruses have been identified. Members of the Coronaviridae are further classified into subgenera. Specifically, in bats, this has included the subgenera, Setracovirus, Myotacovirus, Rhinacovirus, Colacovirus, Pedacovirus, Decacovirus, Minunacovirus, and Nyctocovirus, from the Alphacoronavirus genus, in addition to Nobecovirus, Sarbecovirus, Merbecovirus and Hibecovirus from the Betacoronavirus genus. Coronaviruses from rodents, fall within two subgenera: the Luchacoviruses (within the Alphacoronavirus genus) and Embecoviruses (Betacoronavirus genus) [5] (Table 1).

Table 1

Coronaviridae subgenera associated with bats and rodents

	Order Chiroptera	Order Rodentia
α CoV	ColacovirusDecacovirusDuvinacovirusMinunacovirusMyotacovirusNyctocovirusPedacovirusRhinacovirusSettracovirus	Luchacovirus
β CoV	HibecovirusMerbecovirusNobecovirusSarbecovirus	Embecovirus

Current classification schemes for the Coronaviridae utilize the replicase domain of the ORF1ab gene; however, the spike protein, is a critical mediator of viral tropism and an important contributor to the viruses natural history and potential for spillover [6]. The rodent Alphacoronaviruses have previously been shown to form a monophyletic group [7]. The Embecovirus subgenera, previously classified as lineage 2a, additionally includes human coronavirus HKU-1, several mammalian coronaviruses such as equine coronavirus and dromedary camel coronavirus HKU23 [8], but lacks any representation of Chiroptera-associated coronaviruses [4]. With discovery of the China Rattus coronavirus (ChRCoV) HKU24, Lau and colleagues have provided evidence that rodent coronaviruses are the ancestors to the Embecovirus subgenus [4,9]. While bats are frequently implicated as the source of new and emerging coronaviruses, rodents need also to be considered as a potential source of zoonotic spillovers, given previous observations. The natural ability of coronaviruses to recombine adds to the emergence and transmission of new viruses and there are many examples of recombination reported: the prototypic coronavirus mouse hepatitis virus (MHV), an Embecovirus, has been noted for its ability to recombine [10,11] and the dromedary camel coronavirus HKU23 appears to have had a recombinant history, based on similarity between portions of HKU23 and RodentCoV-IM2014 [12].

Globally, over 2000 rodent species have been described and encompass a wide range of habitats [13]. Rodents have previously been implicated in transmitting numerous pathogens to humans, including hantavirus, Lassa virus, Francisella tularensis and Yersinia pestis. In a study involving rodents destined for human consumption, coronaviruses were frequently identified [14]. Among some of the rodents sampled, bat coronavirus 512/2005 and infectious bronchitis virus were identified; although this may not be an indication of active infection, as the authors note, it does provide evidence of ongoing transmission opportunities across animal species [14]. Two coronaviruses of human consequence have been hypothesized to have rodent origins: HCoV-OC43 and HCoV-HKU1, with HCoV-OC43 emerging via a bovine intermediate [15]. Surveillance efforts have recently identified OC-43, and sequences that cluster with OC-43, in a small number of free-living rodent species [16,17]. In addition to zoonotic rodent transmission concerns, Yang and colleagues have hypothesized a potential role of rodents in the emergence of swine acute diarrhea syndrome coronavirus (SADS-CoV) [18]. Other hypotheses regarding SADS-CoV suggest it may have origins with the bat coronavirus HKU2, which, based on the replicase gene, has previously been classified as a Rhinacovirus, genus Alphacoronavirus, despite the spike protein resembling that of Betacoronaviridae [19]. A virus identified in Apodemus chevrieri, AcCoV-JC34 has been considered to have a unique spike protein compared with other alphacoronaviruses, being only 1126 residues and phylogenetically clustering with the bat coronavirus HKU-2 [20]. Coronaviruses in rodents have been identified across several countries and from a number of rodent host species, including Bandicota savilei, Rattus exulans, Rattus tanezumi, Rattus norvegicus, Rattus argentiventer, Leopoldamys neilli, Rattus andamanensis, Rattus losea, Maxomys surifer, Mus cervicolor, Berylmys berdmorei, Mus caroli, Mus cookie, Niviventer fulvescens, Berylmys bowersi, Dremomys refigenis, and Menetes berdmorei [16,21] (Table 2 and Supplementary Table 1).

Table 2

Surveillance studies investigating coronavirus shedding across rodent species.

Species	Location	Time period	# Positive/Sample size	Ref
Apodemus sp.	France	2014–2016	16/206	[78]
Apodemus agrarius	China	2011–2013	10/444	[5]
	China	2014–2015	1/49	[17]
Apodemus chevrieri	China	2011	21/98	[20]
	China	2014–2015	24/193	[17]
Apodemus ilex	China	2011	1/17	[20]
Apodemus latronum	China	2014–2015	3/6	[17]
Apodemus sylvaticus	United Kingdom	Not given	0/9	[79]
Arvicola terrestris	France	2014–2016	0/35	[78]
Bandicota indica	China	2014–2015	2/5	[17]
Deomys ferrugineus	Democratic Republic of Congo, Republic of Congo	2006–2018	1/1	[80]
Eothenomys cachinus	China	2014–2015	1/1	[17]
Eothenmys fidelis	China	2011	1/62	[20]
Eothenomys miletus	China	2014–2015	3/131	[17]
Field rat (Rattus sp. and Bandicota sp.)	Viet Nam	2013–2014	239/702⁎⁎	[14]
Hystrix sp.	Viet Nam	2013–2014	20/331⁎⁎⁎	[14]
Malacomys longipes	Democratic Republic of Congo, Republic of Congo	2006–2018	1/38	[80]
Micromys minutus	China	2011–2013	0/2	[5]
Microtus sp.	France	2014–2016	0/9	[78]
Microtus agrestis	United Kingdom	Not given	3/11	[79]
Microtus fortis	China	2011–2013	0/305	[5]
	China	2014–2015	0/10	[17]
Mus musculus	China	2011–2013	0/7	[5]
	United Kingdom	Not given	0/394 (Liver)	[79]
			0/58 (Gut)
	China	2014–2015	0/1	[17]
Myodes glareolus	United Kingdom	Not given	1/1	[79]
	Poland	Not given	1/300	[79]
	France	2014–2016	5/80	[78]
Niviventer confucianus	China	2011–2013	1/85	[5]
Niviventer eha	China	2014–2015	0/2	[17]
Niviventer fulvescens	China	2010–2012	0/97⁎⁎⁎⁎	[9]
Niviventer niviventer	China	2014–2015	0/2	[17]
Rattus sp.	Viet Nam	2013–2014	1/1⁎⁎⁎	[14]
Rattus andamanensis	China	2010–2012	0/170⁎⁎⁎⁎	[9]
	China	2014–2015	1/1	[17]
Rattus edwardsi	China	2011–2013	0/2	[5]
Rattus lossea	China	2011–2013	14/301	[5]
Rattus losea	China	2014–2015	0/15	[17]
Rattus fulvescens	China	2011–2013	0/4	[5]
Rattus nitidus	China	2014–2015	0/2	[17]
Rattus norvegicus	United Kingdom	Not given	4/95⁎ (Liver)	[79]
	China	2010–2012	3/359⁎⁎⁎⁎	[9]
	China	2011–2013	4/262	[5]
	China	2014–2015	2/101	[17]
Rattus tanezumi	China	2010–2012	0/9⁎⁎⁎⁎	[9]
	China	2011–2013	1/53	[5]
	China	2014–2015	2/159	[17]
Rattus rattus	China	2010–2012	0/24⁎⁎⁎⁎	[9]
Rattus rattus sladeni	China	2014–2015	0/18	[17]
Rhizomys sp.	Viet Nam	2013–2014	6/96⁎⁎⁎	[14]
Sciuridae sp.	Viet Nam	2013–2014	0/1	[14]

Authors noted that some animals had both liver and gut samples positive (2).

26 of the animals were co-infected; viruses identified were murine coronavirus and Longquan Aa coronavirus.

Virus identified were bat coronavirus 512/2005 or infectious bronchitis virus.

Specifically tested for HKU24 via RT-PCR.

The spillover of coronaviruses from animal species to humans has the potential for major global health implications. Following the 2002–2003 severe acute respiratory syndrome coronavirus (SARS-CoV) outbreak, surveillance studies began to recognize the role that bat species play as reservoirs for novel coronaviruses [22,23]. Ten years later, Middle East respiratory coronavirus (MERS-CoV) emerged in humans and clusters phylogenetically with two bat viruses, Tylonycteris bat coronavirus HKU4 (BtCoV-HKU4) and Pipistrellus bat coronavirus HKU5 (BtCoV-HKU5) [24] in the Merbecovirus clade (group 2c coronaviruses). Further investigation identified NeoCoV from South African Neoromicia capensis bats and suggested it as the sister group to the MERS-CoV-camel clade [25]. MERS-CoV involved the dromedary camel as an intermediate host in the spillover to humans [[26], [27], [28]]. The most recent SARS-CoV-2 outbreak has been traced to bat origins, with a high similarity to a previously identified bat coronavirus RaTG13 (96% genome-wide sequence identity) [29], as well as several other bat coronaviruses, RpYN06 (94.5% sequence identity) [30] and bat-SL-CoVZC45 and bat-SL-CoVZXC21 [31]. The role of an intermediate host in the spread of SARS-CoV-2 to humans remains an open area of investigation.

A major factor for understanding coronavirus tropism lies with the spike protein, consisting of S1 and S2 domains that carry out the receptor binding and membrane fusion functions of the virus, respectively [1]. SARS-CoV-2 and SARS-CoV both utilize ACE-2 as their receptor, while MERS-CoV utilizes DPP4 [[32], [33], [34]]. Equally important as a usable receptor is the presence of a functional protease for S activation [35]. Numerous proteases have been investigated across the Coronaviridae, including TMPRSS2, which acts to activate SARS-CoV-2, as well as furin, trypsin, and cathepsins [36]. While receptor binding is well recognized as a factor in host tropism, it is becoming more recognized that host cell proteases can also influence coronavirus species susceptibility and “spillover” events [37,38].

A notable difference between SARS-CoV-2 and RaTG13 (and all other Sarbecoviruses) is the insertion of a predicted cleavage site for a furin-like protease (P-R-R-A-R) between the S1/S2 regions of the spike protein of SARS-CoV-2 [[39], [40], [41]] with loss of this site associated with virus attenuation and decreased tramnsmission [42,43]. However, the P-R-R-A-R motif is sub-optimal for furin cleavage, indicating it may not be fully adapted for humans. RmYN02, identified in R. malayanus has 93.3% genomic nucleotide identity with SARS-CoV-2 and while the identity is only 71.9% (nucleotide sequence) in the spike protein, RmYN02 has a closely homologous sequence (P-A-A-R) at the S1/S2 cleavage site; however, this is lacking basic amino acids needed for cleavage by furin-like proteases [44]. Comparatively, MERS-CoV is highly unusual in that it contains both an S1/S2 furin cleavage site and an S2’ furin cleavage site [35]. Among two closely related bat coronaviruses, BatCoV-HKU5 and BatCoV-HKU4, HKU5 is also cleaved by furin, though HKU4 may be more closely related to MERS-CoV and can utilize the same receptor, DPP4 [45,46].

Furin is a ubiquitously expressed serine protease that functions across numerous physiological processes [47]. In addition to several other proprotein convertases, furin has the ability to cleave and activate viral proteins [48]. Differences in furin-like activity across animal hosts may impact viral processing or alter viral tropism. Furin cleavage sites are not unique to the Coronaviridae. The highly pathogenic influenza strain, H5N1, for instance, also contains a polybasic furin cleavage site in the hemagglutinin H5 protein, with major implications for disease outcome [49]. Other viral proteins that contain furin cleavage sites include human immunodeficiency virus (HIV) envelope glycoprotein gp160, herpesvirus glycoprotein B, tick-borne encephalitis virus envelope protein prM, and Ebola virus GP [47,50,51]. Additionally, within the Nidovirales (the order within which the Coronaviridae are placed), a furin cleavage site has been predicted in an insect nidovirus isolated from Culex mosquitoes [52].

In the case of coronaviruses, it is known that furin cleavage sites often occur naturally and that this is highly dependent on the specific virus family or lineage [57]. While uncommon for coronaviruses of bats, furin cleavage sites are commonly found in coronaviruses of rodents and it is perhaps fitting to note that proteolytic processing of the coronavirus spike protein was first recognized in the model rodent coronavirus, murine hepatitis virus, MHV-A59 [53], with later analyses demonstrating the importance of furin for the proteolytic cleavage and function of its spike protein [54]. Here, we provide a comprehensive analysis of furin cleavage sites found in rodent coronaviruses, with a focus on the Embecovirus subgenus within the Betacoronavirus family, and discuss implications for virus evolution and zoonotic transfer from rodent species from a One Health perspective.

Furin cleavage across bat and rodent species

Based on the known ability of bats and rodents to harbor coronaviruses, comparing the spike protein and furin cleavage sites of viruses across these two hosts may help elucidate origins of novel coronaviruses and guide future surveillance efforts. Through the use of two publicly available programs used to accurately predict furin cleavage sites, PiTou and ProP [55,56], we analyzed furin cleavage across spike sequences in rodent and chiropteran species. PiTou utilizes a hidden Markov model, specifically targeting 20 amino acid residues surrounding furin cleavage sites and important for binding and solvent accessibility [56]. The final score in PiTou is based on log-odds probability. ProP utilizes an artificial neural network to predict furin cleavage [55]. In ProP, sites are given a score of between 0 and 1, with a score above 0.5 being a predicted furin cleavage site.

The protease furin cleaves at a distinct multi-basic motif containing paired arginine residues; furin requires a minimal motif of R-X-X-R, with a preference for an addition basic residue; i.e., R-X-B-R [48]. While most studies have focused on human furin, previous work has indicated furin-like proteases in the megabat Pteropus [57]. The presence of suitable protease activators to enable viral infections further adds to the mystery that enables bats to act as viral reservoirs without seemingly showing clinical signs. Many, but not all, coronaviruses contain a cleavage site at S1/S2, which primes the spike protein for fusion and may increase its affinity for the viral receptor but may also make it structurally unstable. As such S1/S2 is considered dispensable for virus infection. The presence of S1/S2 furin cleavage sites in viruses uniquely identified in bats is shown in Fig. 1A (see also Supplementary Table 2). Across the Coronaviridae, the S2’ cleavage site is also present, with cleavage here considered a required event for viral infection [58]. Bat sequences which have predicted furin cleavage sites at S2’ are additionally shown in Fig. 1B.

Fig. 1

Examples of predicted furin cleavage sites identified in coronaviruses associated with chiropteran species. Over 150 spike sequences from bat associated CoVs were screened for furin cleavage sites, including those in both the Alpha- and Betacoronavirus genera, using the programs PiTou and ProP. A. Unique S1/S2 furin cleavage sites predicted in bat associated coronaviruses. B. Unique S2’ furin cleavage sites predicted across bat associated coronaviruses. Associated NCBI accession numbers are as follows: Bat Hp-betacoronavirus/Zhejiang2013 (YP_009072440), (Putative) Zaria bat coronavirus (ADY17911), Bat coronavirus HKU5–1 (ABN10875), BtPa-BetaCoV/GD2013 (AIA62343), Pipistrellus abramus bat coronavirus HKU5-related (QHA24687), Bat coronavirus HKU5-2 (ABN10884), Pipistrellus bat coronavirus HKU5 (AGP04938), Coronavirus Neoromicia/PML-PHE1/RSA/2011 (AGY29650), BtCoV/A434/2005 (ABG11962), Bat coronavirus PREDICT/PDF-2180 (YP_009361857), BtRt-BetaCoV/GX2018 (QJX58383), Rousettus bat coronavirus HKU 9 (AVP25406), Bat coronavirus 1B (ACA52157), 229E-related bat coronavirus (ALK28781), BtMf AlphaCoV/GD2012-b (AIA62241), Bat coronavirus MfulBtCoV/3709 (AMB43191), Bat coronavirus MfulBtCoV/3736-1 (AMB43194), Bat coronavirus MfulBtCoV/3759-1 (AMB43195), Bat coronavirus MsBtCoV/4001-1 (AMB43196), Bat coronavirus MsBtCoV/4068 (AMB43198), Rhinolophus bat coronavirus HKU32 (QCX35178), Rousettus bat coronavirus GCCDC1 (QKF94914), Hipposideros bat coronavirus HKU10 (AFU92131), Bat coronavirus HKU9-3 (ABN10927), Bat coronavirus HKU9-1 (ABN10911), Alphacoronavirus Bat-CoV/P.kuhlii/Italy/3398-19/2015 (YP_009755890), NL63-related bat coronavirus (YP_009824967), Chaerephon bat coronavirus/Kenya/KY41/2006 (ADX59458), Miniopterus bat coronavirus/Kenya/KY33/2006 (ADX59488), SARS-like coronavirus BatCoV/BB9904/BGR/2008 (ALJ94036), and Rousettus bat coronavirus/Kenya/KY06/2006 (ADX59474).

The S1/S2 cleavage site was first noticed in coronaviruses infecting laboratory rodents; for example, the mouse hepatitis virus (MHV-1) strain JHM contains a strong S1/S2 cleavage site in addition to a second predicted furin cleavage site just distal. The presence of S1/S2 furin cleavage sites in viruses uniquely identified in rodents is shown in Fig. 2A. Most often, a second, overlapping, S1/S2 furin cleavage site was predicted by PiTou and not by ProP, though an exception was rodent coronavirus RtBi-CoV/FJ2015, in which both PiTou and ProP predicted adjacent furin cleavage sites. This redundancy may indicate a needed feature for rodent-associated Coronaviridae. Unsurprisingly, the majority of the rodent coronaviruses with S1/S2 furin cleavage sites were classified as Embecoviruses, a subgenera known for this feature [59]. A weakly predicted cleavage site was also detected in the Alphacoronavirus, AcCoV-JC34. A total of 21 unique potential S1/S2 furin cleavage sites are shown in Fig. 2A. It is important to note, however, that isolate specific differences exist; for example, two isolates of HKU24 possess disparate predicted furin cleavage sites. Lastly, the virus with the strongest predicted furin cleavage site, based on PiTou, was RtR-CoV-T12006A. ProP also gave a strong furin cleavage prediction score (0.88), the highest score observed from ProP, but shared by several other viruses. S2’ furin cleavage sites additionally occur naturally in rodent coronaviruses and are shown in Fig. 2B. It is worthwhile to note, the rodent coronaviruses with S2’ furin cleavage sites also possess S1/S2 furin cleavage sites. Compared to S1/S2 furin cleavage sites in rodent coronaviruses, S2’ furin cleavage sites are less common. Given the well-conserved nature of the S2’ cleavage site, it is unsurprising that the predicted scores in the rodent coronaviruses are relatively similar to the predicted scores from bat associated coronaviruses.

Fig. 2

S1/S2 furin cleavage sites are commonly found in rodent associated coronaviruses. A. Examples of predicted S1/S2 furin cleavage sites in rodent associated coronaviruses. Longquan RI rat coronavirus Longquan-189 (AID16649) shared the same S1/S2 furin cleavage site as Longquan Rl rat coronavirus Longquan-370 (AID16655). Murine hepatitis virus strain JHM (YP_209233) shares the same S1/S2 furin cleavage sites as several other MHV strains (ACN89763, ACN89705, ACN89696, ACN89722), as well as Murine coronavirus MHV-3 (ACN89743) in which sites occur at residues 770 and 769, versus 769 and 768. Longquan Rl rat coronavirus Longquan-708 (AID16661) shares the same S1/S2 cleavage site as Murine coronavirus MHV-1 (ACN89742), in which the predicted furin cleavage site occurs at residue 758. Betacoronavirus HKU24 Lijiang-41 (QOE77297) shares the same S1/S2 furin cleavage site as Betacoronavirus HKU24 Lijiang-53 (QOE77307) and Betacoronavirus HKU24 Ruili-874 (QOE77327), in addition to several other accession numbers (AYR18625, AYR18679, AYR18634, AYR18607, AYR18652, AYR18670, and AYR18643). Coronavirus AcCoV-JC34 (YP_009380521) shares the same S1/S2 furin cleavage site as Lucheng Rn rat coronavirus Lijiang-170 (QOE77268), which were only predicted by PiTou. RtRt-CoV/Tb2018 (QIM73854) shared the same S1/S2 predicted cleavage sites as RtRt-CoV/Tk2011 (QIM73813). Rodent coronavirus RtMm-CoV/GD2015 (ATP66756) shared the same S1/S2 predicted cleavage site as Rodent coronavirus RtRn-CoV/YN2013 (ATP66727). RtRe-CoV/Tl2009 (QIM73841) shared the same S1/S2 furin cleavage site as RtRt-CoV/Tn2018 (QIM73848). B. S2’ furin cleavage sites predicted in rodent association coronaviruses, both of which were noted as unclassified, but our analysis supported as belonging to the Embecovirus subgenera. *Our analysis supports classification as an Embecovirus.

Phylogenetic analysis

The majority of rodent coronaviruses investigated here are from the Embecovirus clade. This clade additionally includes two human coronaviruses, HCoV-OC43 and HCoV-HKU1 [15]. As previously reported and supported by our analysis, HCoV-HKU1 likely emerged through a rodent host [15]. Specifically, the phylogenetic history of the spike protein of Embecoviruses is shown in Fig. 3A. The tree is rooted at the HKU24 group based on earlier reports [9] as well as our own findings, which involved rooting the Embecovirus clade with three different Merbecovirus CoV – including those from a hedgehog and two bats – using a 2000 amino acid alignment after Gblocks [60] removal of ambiguously aligned positions of Orf1ab, and in all cases HKU24 fell at the base of the Embecovirus group. The phylogeny depicts a basal split in the Embecoviruses into a predominately rodent clade and a predominately ungulate clade, with spike sequences from horses at the base of that ungulate group. Historically, rodents have been considered major sources of pathogens and our evidence support their role in the emergence of two human coronaviruses (HCoV-OC43 and HCoV-HKU1).

Fig. 3

Phylogenetic tree of spike protein sequences from the Embecovirus clade. A) The tree was constructed using RaxML [81], 1000 bootstrap replicates, from a spike protein alignment, that removed any ambiguously aligned positions using Gblocks [60]. Bootstrap support percentages for nodes receiving less than 75% are not shown. B) Sequences are colored based on the predicted furin cleavage score from PiTou.

Furin scores using both ProP and PiTou are summarized in Fig. 3B, based on PiTou scoring for all the Embecoviruses spike genes analyzed in Fig. 3. Furin scores are represented in a heat map coloring, with lower score in red and higher score in blue. Unscored residues are depicted with a gray box.

Almost all Embecoviruses contain furin cleavage sites at the S1/S2 interface, with two notable exceptions (shown with black boxes in Fig. 3B). MHV-2 (AF201929) is the cause of a laboratory-acquired infection, first identified in 1952 as acute hepatitis associated with mouse leukemia [61]. It is a distinct virus in cell culture, with a spike protein known to be cleaved by cathepsins [62] and differs significantly from the more widely studied MHV-I and may represent a tissue-specific variant (S1/S2 site HRARSS for MHV-2 vs. HRARRS for MHV-1; basic residues in bold, predicted mutation underlined). As such, MHV-2 may show parallels with a tropism shift associated with a loss of furin cleavage seen with feline coronaviruses [63]. The other notable exception is Longquan Aa mouse coronavirus (LAMV) (AID16631). Comparatively HKU24 (ATP66750) contains an S1/S2 furin cleavage and aligns with specific residues LL-GAPRE for LAMV. In this case, the S1/S2 site seems to represent an indel at that region, and would be of high interest for further follow up with regard to its evolution.

In some cases, there is discrepancy between the ProP and PiTou scores, notably for sequences YP_009113025 and QEY10649. YP_009113025 is a variant of HKU24 with an S1/S2 site of WRARRD, with a bulky hydrophobic residue (W) at P5 and QEY10649 is an HKU23 (dromedary camel) virus with an S1/S2 site DRRxRRA is a with a negatively-charged/acidic residue (D) at P6. In both cases, we consider that these flanking residues are the cause of the scoring discrepancy, with PiTou downgrading the score because of this. The impact of such flanking residues with altered biochemical properties remains to be resolved, but equivalent changes have also been seen in modified furin cleavage observed with feline coronaviruses, as measured with enzymatic assays with purified proteases and peptide mimics of the S1/S2 cleavage site [63]. Overall, while a furin cleavage site is prevalent in Embecovirus spike protein, the composition of the site does vary and a more detailed study of these variations may inform our understanding of infection and zoonotic transfer of these viruses.

Discussion

Continued surveillance in bat and rodent species, concomitant with laboratory experimentation and computational analysis, is essential for understanding the overall disease ecology of coronaviruses in these two host groups. Understanding species cross-over potential for coronaviruses often focuses on known features, such as mutation within the receptor-binding domains or in genomic recombination events, but other factors, such as alterations in protease cleavage sites within spike, may also be important.

The Chiroptera and Rodentia encompass vast lineages and inhabit unique ecological niches from bats roosting in caves to mice living in apartment buildings. The spillover of coronaviruses between these species and into humans, livestock, and other species remains a threat to One Health and global health. Across species, it is also important to consider host adaptations that may facilitate their ability to act as viral reservoirs. In bats, the ability to harbor and shed coronaviruses, with limited or inapparent clinical signs has remained an ongoing question [38,64]. By comparison, less has been investigated in wild rodent species harboring coronaviruses, though in laboratory settings it is apparent that rodents can show disease associated with coronavirus infection, including Golden Syrian Hamsters (Mesocricetus auratus) infected with SARS-CoV-2, which have proven a robust and effective model for SARS-CoV-2 infection [65,66].

SARS-CoV-2 has remained a global challenge and like most diseases, rodent models are helpful in guiding our understanding. In laboratory mice, transgenic humanized ACE2 mice better recapitulate SARS-CoV-2 infection and pathology [67,68]. However, new SARS-CoV-2 variants of concern (VOCs) have expanded the host range of SARS-CoV-2, allowing infection of laboratory mice [69]. Challenges of deer mice (Peromyscys maniculatus) with SARS-CoV-2 have demonstrated infection and transmission [70,71]. Additionally, experimentally challenged bank voles (Myodes glareolus) appear susceptible to SARS-CoV-2, despite not showing clinical signs [72]. Understanding coronavirus adaptation and transmission among rodent species is an important area of investigation both from the One Health perspective as well as understanding spillover between rodents. Additional important areas of study include how/why some rodents harbor unique viruses, as well as what are the underlying host differences that previously allowed one rodent species (laboratory mice) to be largely unaffected by SARS-CoV-2 while another rodent species (hamsters) readily develop disease. In mice infected with MHV, disease has been associated with encephalomyelitis, wasting, and mortality in naïve animals [73,74]. Further, a case report has described a wasting syndrome in guinea pigs with a presumed coronavirus infection [75].

The activation of the spike protein is a complex process and furin is a commonly used protease to activate fusion machinery. Among the Alphacoronaviruses, the presence of an S1/S2 furin cleavage site is infrequently observed, though a notable example is in feline coronavirus, which normally possesses a furin cleavage site at the S1/S2 boundary, but loss of basic residues is often associated with the systemic disease feline infectious peritonitis [63]. A second Alphacoronavirus, canine coronavirus 23/03, also possesses a furin cleavage site, but the role in disease is less well defined [76]. In bat spike sequences from Alphacoronaviruses, no obvious S1/S2 furin cleavage site was identified based on ProP and PiTou predictions. Two proposed alphacoronavirus from rodents, AcCoV-JC34 and Lucheng Rn rat coronavirus Lijiang-170, had a shared, weakly predicted furin cleavage site (S-R-R-A-R), based on the PiTou program. A caveat to using bioinformatics programs lies in the correlation between a software prediction and biological plausibility. A predicted site, for example, may not be accessible to furin and thus, non-functional and the action of other proteases, needs also to be considered. It is also presently unclear how the presence or absence of a cleavage site and its relative cleavability score translates into a given biological process, such a transmissibility, cross-species tropism or pathogenicity. Nevertheless, our study provides a predictor that can be integrated with experimental validation and further surveillance.

Based on our studies reported here, it is possible that viruses with S1/S2 furin cleavage sites are more commonly found in Coronaviridae circulating in rodents. Over three times as many spike sequences were publicly available and screened from bats (179) compared to rodents (55). While MHV was included in our screening, 41 rodent viruses were from previous surveillance work and can be considered naturally occurring. Of these 41 rodent sequences, 32 (78%) had potential S1/S2 furin cleavage sites. In the spike sequences from bats, only 11 of 179 (6%) sequences had predicted S1/S2 furin cleavage sites. For some CoVs, such as HKU5, numerous accessions are publicly available through sources such as NCBI, so it is challenging to strictly compare the frequency of furin cleavage sites across the species that sequences come from, but it does seem apparent that rodents regularly harbor viruses with strong S1/S2 furin cleavage sites. With regard to the S2’ furin cleavage site, the proportion of rodent versus bat sequences possessing this site is relatively similar, being found in 3 of 41 (7.3%) naturally occurring rodent spike sequences and in 13 of 179 (7.3%) bat sequences. Host differences may additionally contribute to these observed differences. Both rodent and bat associated coronaviruses have been associated with human disease and both must continue to be investigated, including in regular surveillance studies. Recently, a web-based, risk assessment tool for evaluating viruses of wildlife origin designated Lassa virus, a rodent-associated virus, as the virus most likely to spillover into human populations (ranked #1); notably, several rodent-related coronaviruses were ranked in the top 50, specifically Murine CoV (#18), Longquan Aa CoV (#23), Rodent CoV (#27), and HKU-1 (37) [77].

The Coronaviridae lend themselves well to zoonotic spillover. The unique spike protein determines host tropism and is largely a balance between receptor binding ability and the presence of an acceptable host cell protease for spike protein activation. Bats and rodents have both been implicated as reservoirs for ancestral coronavirus species. The common presence of predicted furin cleavage sites across rodent coronaviruses, suggests rodents warrant further consideration as a putative source of new viral emergences.