Benign Rabbit Caliciviruses Exhibit Evolutionary Dynamics Similar to Those of Their Virulent Relatives

ABSTRACT

Two closely related caliciviruses cocirculate in Australia: rabbit hemorrhagic disease virus (RHDV) and rabbit calicivirus Australia 1 (RCV-A1). RCV-A1 causes benign enteric infections in the European rabbit (Oryctolagus cuniculus) in Australia and New Zealand, while its close relative RHDV causes a highly pathogenic infection of the liver in the same host. The comparison of these viruses provides important information on the nature and trajectory of virulence evolution, particularly as highly virulent strains of RHDV may have evolved from nonpathogenic ancestors such as RCV-A1. To determine the evolution of RCV-A1 we sequenced the full-length genomes of 44 RCV-A1 samples isolated from healthy rabbits and compared key evolutionary parameters to those of its virulent relative, RHDV. Despite their marked differences in pathogenicity and tissue tropism, RCV-A1 and RHDV have evolved in a very similar manner. Both viruses have evolved at broadly similar rates, suggesting that their dynamics are largely shaped by high background mutation rates, and both exhibit occasional recombination and an evolutionary environment dominated by purifying selection. In addition, our comparative analysis revealed that there have been multiple changes in both virulence and tissue tropism in the evolutionary history of these and related viruses. Finally, these new genomic data suggest that either RCV-A1 was introduced into Australia after the introduction of myxoma virus as a biocontrol agent in 1950 or there was drastic reduction of the rabbit population, and hence of RCV-A1 genetic diversity, perhaps coincident with the emergence of myxoma virus.

IMPORTANCE The comparison of closely related viruses that differ profoundly in propensity to cause disease in their hosts offers a powerful opportunity to reveal the causes of changes in virulence and to study how such changes alter the evolutionary dynamics of these pathogens. Here we describe such a novel comparison involving two closely related RNA viruses that cocirculate in Australia, the highly virulent rabbit hemorrhagic disease virus (RHDV) and the nonpathogenic rabbit calicivirus Australia 1 (RCV-A1). Both viruses infect the European rabbit, but they differ in virulence, tissue tropism, and mechanisms of transmission. Surprisingly, and despite these fundamental differences, RCV-A1 and RHDV have evolved at very similar (high) rates and with strong purifying selection. Furthermore, candidate key mutations were identified that may play a role in virulence and/or tissue tropism and therefore warrant further investigation.

INTRODUCTION

Revealing the factors that determine the evolution of virulence in viruses and identifying how changes in virulence might impact viral evolutionary dynamics are key queries in the study of infectious disease evolution. Although there is a large body of work on revealing the relationship between virulence and transmissibility, most studies of virulence evolution involve the retrospective analysis of single viruses (1–4). However, the comparison of closely related viruses that differ profoundly in virulence would provide a powerful opportunity to reveal the trajectory of virulence evolution, its determinants, and its impact on other aspects of viral evolution. The evolutionary dynamics of the highly virulent rabbit hemorrhagic disease virus (RHDV) have recently been characterized (5, 6). However, similar studies of closely related, nonpathogenic rabbit caliciviruses are lacking, even though they circulate in the same populations and may provide important insights into virulence evolution. Here we describe, for the first time, the detailed evolutionary analysis of such a benign relative, the nonpathogenic rabbit calicivirus Australia 1 (RCV-A1), that is cocirculating with virulent RHDV in Australia.

RHDV and RCV-A1 represent different lineages within the genus Lagovirus of the family Caliciviridae (7–10). RCV-A1 and RHDV share many genomic features, such as a polyadenylated single-stranded positive-sense RNA genome of around 7.5 kb organized into two overlapping open reading frames (ORFs) with a viral protein (VPg) covalently bound to the 5′ end (Fig. 1) (10–12). The two viruses also have an identical genome organization, in which ORF 1 is divided into genes coding for seven nonstructural proteins (including a helicase, protease, and RNA-dependent RNA polymerase [RdRp]) and the major capsid protein (VP60), while ORF 2 encodes a minor structural protein (VP10) (Fig. 1) (10, 11, 13). The capsid protein comprises three domains: the N-terminal arm, the conserved shell domain, and the protruding domain, which is further classified into the P1 and P2 subdomains (14). P2 is the least conserved domain, as it is the most exposed and likely contains determinants for antigenicity and host cell interactions (14).

FIG 1.

Schematic representation of the RHDV and RCV-A1 genomes. (Top) Both genomes are organized into two ORFs (open boxes), with short 5′ and 3′ untranslated regions (UTRs) (lines), and are VpG linked and polyadenylated. (Bottom) ORF 1 encodes a polyprotein that is proteolytically cleaved into multiple proteins (open boxes), including the 2C-like helicase, the protease, RdRp, and VP60 (capsid protein). As indicated and separated by dotted lines, VP60 has three structural domains, the N-terminal arm (N), the shell domain (S), and the protruding (P) domain, which is split into P1 and P2 subdomains. ORF 2 encodes a minor structural protein, VP10.

The highly virulent RHDV causes acute hemorrhagic disease in the European rabbit (Oryctolagus cuniculus), targeting primarily the liver and spleen (15) with a case fatality rate of >90% (16). It was first documented in domestic rabbits in China in 1984 (17), from where it spread rapidly, causing outbreaks in domestic and wild rabbits in multiple continents and eventually becoming endemic in wild rabbit populations in Europe (18–21). RHDV was deliberately introduced as a biological control agent into both Australia and New Zealand in the mid-1990s to suppress European rabbit populations, which are responsible for considerable agricultural and ecological damage (22–24). Prior to that, myxoma virus (MYXV) had been introduced into Australia in the 1950s to combat the ever-increasing rabbit population and was initially extremely effective (25, 26). However, host-pathogen coevolution led to a combination of MYXV attenuation and mounting host resistance in Australian rabbits, which allowed a degree of reexpansion of the rabbit population (27).

RCV-A1 differs from RHDV in both tissue tropism and virulence. RCV-A1 was first isolated from a wild rabbit in Australia (10) and causes a mild localized infection in the small intestine. No clinical signs have ever been associated with this virus (28, 29). Nonpathogenic lagoviruses are not unique to Australia. Several other nonpathogenic lagoviruses have been reported in Europe (RCV-E) (30–32); these were isolated from healthy rabbits and experimentally demonstrated to cause a similarly benign enteric infection (30, 31). RCV-A1 was also recently detected in New Zealand (L. J. Nicholson, J. E. Mahar, T. Strive, V. Ward, and J. Duckworth, unpublished data).

Both nonpathogenic and pathogenic lagoviruses cocirculate in wild and domestic rabbit populations in Europe, Australia, and New Zealand. Due to their genetic and antigenic similarities, nonpathogenic lagoviruses can provide immunological cross-protection against lethal infection with RHDV; this ranges from complete to no apparent protection (30, 31, 33). Although the cross-protection conveyed by antibodies against RCV-A1 is both partial and transient (28, 33), it is impeding effective rabbit control efforts in Australia. RCV-A1 is found predominantly in the more temperate regions of Australia with higher than average rainfall, where RHDV has proven to be less effective in rabbit control (9, 34–36).

The limited geographical distribution of RCV-A1 compared to RHDV (Fig. 2) (9, 35) can in part be explained by differences in mechanisms of transmission. RCV-A1 (and other nonpathogenic lagoviruses) are likely spread exclusively via the fecal-oral route (10, 35). Although RHDV utilizes the same transmission route, it has also acquired mechanical insect transmission (37, 38), involving both bushfly (Musca vetustissima) and blow fly (Calliphoridae) species in Australia, thereby enabling longer-distance spread between rabbit populations that are not directly connected (24, 39–41). Critically, since rabbit carcasses (rather than diseased animals) are considered the likely source of mechanical insect transmission, this is thought to select for the high case fatality rates observed for RHDV (39, 40), which is always associated with liver tropism.

FIG 2.

Distribution of RHDV and RCV-A1 in Australia and sampling locations. Areas where RHDV occurs in Australia are shaded in pink (based on rabbit distribution) (22, 60, 61). Antibodies to RHDV have been identified in almost every Australian rabbit population tested, such that the distribution of RHDV is likely to be very similar to the distribution of rabbits (22, 60). The distribution of RCV-A1 is limited to the more temperate climate zones in the southeast of the continent and always overlaps with RHDV, as shaded in purple (17). Collection sites for Australian samples sequenced in this study are marked with blue dots and red dots for RCV-A1 and RHDV, respectively. Australian state and territory abbreviations are as follows: WA, Western Australia; NT, Northern Territory; SA, South Australia; QLD, Queensland; NSW, New South Wales; ACT, Australian Capital Territory; VIC, Victoria; and TAS, Tasmania.

The origins of virulent RHDV have been the subject of debate. A recent hypothesis is that the high virulence of RHDV may stem from a cross-species transmission event, resulting from the repeated translocation of New World lagomorphs into Europe (42). A competing theory is that RHDV evolved from an avirulent (i.e., RCV-A1-like) virus within European rabbits without a species jump (24, 30, 43). If true, this would represent a profound change in virulence occurring within a single host species. As all highly virulent forms of RHDV described so far target predominantly the liver, it is tempting to speculate that the acquisition of liver tropism was the key event enabling the emergence of virulent RHDV, particularly as the production of virus-laden carcasses enables faster viral spread over greater distances via carrion-feeding flies.

While all highly virulent RHDV variants known to date target the liver, liver tropism is not always associated with high levels of virulence. For example, a semipathogenic lagovirus (Michigan rabbit calicivirus [MRCV]) isolated in Michigan, USA, reportedly targets the liver but had a lower mortality rate (32.5%) (44, 45). More recently, a new genetically distinct variant, RHDV2 (RHDVb), has emerged; it was first reported in France and Spain in 2010 (46, 47). Interestingly, early reports describe this virus as moderately virulent, with case fatality rates ranging from 0% to 75% in one small study (48). However, when RHDV2 subsequently spread to multiple countries, including Australia (49–54), it caused a notable reduction in rabbit numbers (55). Recombination of RHDV2 with lagoviruses from other lineages (including RCV-A1-like variants) has also been detected (56), which may be of importance as recombination has been proposed as a factor that facilitated the evolution of high virulence in RHDV (57).

Given the potential influence of RCV-A1 on the spread and disease impact of RHDV, it is important to understand the evolutionary history of these cocirculating but profoundly different viruses. It is of particular importance to determine if changes in virulence are associated with differences in evolutionary dynamics and the patterns and signs of natural selection. While the evolution of RHDV has been examined in some detail (6, 41, 43, 58), there are few such parallel studies on nonpathogenic RCVs. An earlier study explored the evolution of RCV-A1 using VP60 gene sequences from viruses sampled over a 2-year period, from which it was suggested that RCV-A1 arrived in Australia with the first rabbits in the mid-19th century (9, 10). However, a small sample size and a limited sampling time frame made it unfeasible to reliably estimate the rate of evolution for RCV-A1, and thus a previously determined (43) (and likely erroneous [6, 59]) evolutionary rate for RHDV was utilized (9). Here, we examined the evolution of RCV-A1 using full genome sequences of strains sampled over a longer (7.5-year) time frame, enabling us to undertake a more detailed comparison between two closely related RNA viruses that differ profoundly in virulence.

MATERIALS AND METHODS

Specimen collection.

Samples of RCV-A1 collected between 2007 and 2009 at Gungahlin (Australian Capital Territory [ACT]), Valpine (New South Wales [NSW]), Oakey Creek (NSW), Cattai National Park (NSW), Michelago, (NSW), Burragate (NSW), Wauchope rabbitry (NSW), Bacchus Marsh (Victoria [VIC]), and Bendigo (VIC) were obtained as part of a previous study (9) and had been stored at −20°C. Additional specimens were collected from these and another three sites between 2010 and 2014 (Table 1) during routine field trips carried out as part of RHDV monitoring. RHDV samples were collected from seven sites in four Australian states and territories: ACT, NSW, South Australia, and Western Australia (Table 1). RCV-A1 samples were collected from the duodena of healthy shot rabbits, while RHDV samples were recovered from the livers of rabbits that were found dead. All methods involving trapping, shooting, and handling of live animals were approved by the CSIRO Sustainable Ecosystems Animal Ethics Committee (SEAEC 06-31, SEAEC 09-14, and SEAEC 12-15) following guidelines of the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. The sample from Gore, Southland, in New Zealand was obtained from the duodenum of a shot rabbit, and sampling was approved by the Landcare Research Animal Ethics Committee (LCRAEC 13/07/02) under the New Zealand Animal Welfare Act 1999 (Part 6, Use of Animals in Research, Testing and Teaching). The sampling locations and general distributions of RCV-A1 (35) and RHDV (22, 60, 61) in Australia are shown in Fig. 2.

TABLE 1.

Collection details for samples used for full-genome sequencing

Statea	Location	Collection date (day/mo/yr)	Name	Species	Accession no.b
ACT	Gudgenby, Canberra	26/11/13	Gudg-26	RCV-A1	KX357655
ACT	Gudgenby, Canberra	8/10/14	Gudg-79	RCV-A1	KX357656
ACT	Gudgenby, Canberra	8/10/14	Gudg-89	RCV-A1	KX357657
ACT	Gungahlin, Canberra	22/02/12	GUN1-240	RCV-A1	KX357658
ACT	Gungahlin, Canberra	15/11/07	GUN1-11	RCV-A1	KX357665
ACT	Gungahlin, Canberra	28/11/08	GUN1-22	RCV-A1	KX357666
ACT	Gungahlin, Canberra	3/12/08	GUN1-29	RCV-A1	KX357659
ACT	Gungahlin, Canberra	19/01/10	GUN1-60	RCV-A1	KX357669
ACT	Gungahlin, Canberra	18/08/10	GUN-121	RCV-A1	KX357660
ACT	Gungahlin, Canberra	31/03/11	GUN-159	RCV-A1	KX357661
ACT	Gungahlin, Canberra	31/03/11	GUN-160	RCV-A1	KX357662
ACT	Gungahlin, Canberra	22/02/12	GUN-238	RCV-A1	KX357663
ACT	Gungahlin, Canberra	11/11/14	GUN-343	RCV-A1	KX357664
NSW	Burragate	17/10/07	BUR1-1	RCV-A1	KX357673
NSW	Cattai National Park, Hawkesbury	6/09/07	CAT 2-12	RCV-A1	KX357674
NSW	Cattai National Park, Hawkesbury	7/09/07	CAT 3-4	RCV-A1	KX357675
NSW	Little Whiskers Rd, Carwoola	20/09/10	LWR-1	RCV-A1	KX357676
NSW	Michelago	23/05/07	MIC3-3	RCV-A1	KX357679
NSW	Michelago	25/09/07	MIC5-8	RCV-A1	KX357680
NSW	Oakey Creek, Bathurst	7/11/07	OC-7	RCV-A1	KX357690
NSW	Oakey Creek, Bathurst	9/11/07	OC-13	RCV-A1	KX357682
NSW	Oakey Creek, Bathurst	9/11/07	OC-15	RCV-A1	KX357683
NSW	Oakey Creek, Bathurst	10/11/07	OC-18	RCV-A1	KX357684
NSW	Oakey Creek, Bathurst	1/12/07	OC-20	RCV-A1	KX357685
NSW	Oakey Creek, Bathurst	1/12/07	OC-26	RCV-A1	KX357686
NSW	Oakey Creek, Bathurst	1/12/07	OC-33	RCV-A1	KX357687
NSW	Oakey Creek, Bathurst	1/12/07	OC-36	RCV-A1	KX357688
NSW	Oakey Creek, Bathurst	1/12/07	OC-39	RCV-A1	KX357689
NSW	Oakey Creek, Bathurst	4/10/11	OAK NT-12	RCV-A1	KX357681
NSW	Valpine, Bathurst	8/11/07	V-4	RCV-A1	KX357691
NSW	Valpine, Bathurst	8/11/07	V-5	RCV-A1	KX357692
NSW	Valpine, Bathurst	8/11/07	V-8	RCV-A1	KX357693
NSW	Wauchope rabbitry	15/05/09	WAU-1	RCV-A1	KX357694
VIC	Bacchus Marsh	31/10/07	BM-2	RCV-A1	KX357701
VIC	Bacchus Marsh	27/08/09	BM-35	RCV-A1	KX357702
VIC	Bacchus Marsh	28/08/09	BM-45	RCV-A1	KX357703
VIC	Bacchus Marsh	28/08/09	BM-49	RCV-A1	KX357704
VIC	Bacchus Marsh	28/08/09	BM-60	RCV-A1	KX357705
VIC	Bendigo	26/08/09	BEN-10	RCV-A1	KX357696
VIC	Bendigo	26/08/09	BEN-35	RCV-A1	KX357699
VIC	Bendigo	6/07/10	BEN-52	RCV-A1	KX357700
VIC	Bendigo	28/11/10	BEN-115	RCV-A1	KX357697
VIC	Bendigo	28/11/10	BEN-124	RCV-A1	KX357698
	Gore, Southland, New Zealand	15/05/13	Gore-425A	RCV-A1	KX357707
ACT	Ainslie, Canberra	1/05/09	AIN-1	RHDV	KX357653
ACT	Ainslie, Canberra	11/06/09	AIN-2	RHDV	KX357654
ACT	Gungahlin, Canberra	10/03/09	GUN1-37	RHDV	KX357667
ACT	Gungahlin, Canberra	13/03/09	GUN1-52	RHDV	KX357668
ACT	Mount Painter, Canberra	2/03/10	MtPt-2	RHDV	KX357670
ACT	Mount Painter, Canberra	2/03/10	MtPt-4	RHDV	KX357671
ACT	Pine Island, Canberra	8/06/09	PI-1	RHDV	KX357672
NSW	Oakey Creek, Bathurst	10/03/07	M2	RHDV	KX357677
NSW	Oakey Creek, Bathurst	12/08/07	M9	RHDV	KX357678
SA	Oraparina, Flinders Ranges National Park	29/09/08	ORA383	RHDV	KX357695
WA	Bakers Hill	30/06/13	B.Hill	RHDV	KX357706

^aAbbreviations for Australian states and territories: ACT, Australian Capital Territory; NSW, New South Wales; SA, South Australia; VIC, Victoria, WA, Western Australia.

^bSequences were submitted to GenBank under the listed accession numbers.

RNA extraction and cDNA synthesis.

Viral RNA was extracted from liver tissue (RHDV) or duodenum (RCV-A1). Tissue (30 mg) was homogenized with glass beads using the Precellys 24 tissue homogenizer, and RNA was extracted using the RNeasy minikit (Qiagen). cDNA was prepared using the Superscript III first-strand synthesis system (Life Technologies), using virus gene-specific primers.

Amplification of viral genomes.

RHDV and RCV-A1 genomes were amplified from cDNA using the Platinum Taq DNA polymerase high-fidelity kit (Life Technologies) according to the manufacturer's instructions. The RHDV genome was amplified either in one fragment using primers RHDV-1 and RHDV-13 (40) or in three overlapping fragments, as described previously (5). RCV-A1 genomes were amplified in six overlapping fragments using primer sets described in Table 2. For difficult templates, additional primer pairs were used to amplify smaller fragments (Table 2). Reaction mixtures were denatured at 94°C for 2 min and then incubated at 94°C for 15 s, 65°C for 30 s (reducing by 0.5°C each cycle), and 68°C for 2 min for 45 cycles, followed by a 5-min incubation at 68°C. PCR products were viewed on 1% agarose gels stained with SYBR safe DNA stain (Life Technologies) before purification using the QIAquick PCR purification kit (Qiagen). Purified PCR fragments for each sample were pooled at equimolar concentrations and diluted in nuclease-free water for use as input DNA for library preparation.

TABLE 2.

RCV-A1 amplification primers

Fragment and primer name	Sequence (5′→3′)	Sense	Position (nt)	Size (kb)	Reference
1
MICV12	GTG AAA GTT ATG GCG GTT TTA TCG	+	1–24	1,786	10
RCVr1.8	CTT TGT CAC AGT TGA GRG GGC A	−	1765–1786		This study
2
RCVf1.5	CAG CCY GTT GCY GTC ATY TTC AA	+	1537–1559	1,789	This study
RCVr3.3	GGR AGY CCY TCA TAG TCA TTG TCA T	−	3302–3326		This study
3
RCVf3.0	GGY AAT GAY GAG TAT GAY GAG TGG CA	+	3021–3047	1,656	This study
RCVr4.7	ATR CCA CTT GGR AGY CCT CTT TTR G	−	4652–4676		This study
4.2
Lago9	TGG NCC NAT YGC AGT YGG VRT TGA CAT GAC	+	4401–4430	1,674	9
RCVr6.1	ACT ATC TGR CCR TTC CAY CTG TTG TC	−	6049–6074		This study
5a
RCVf5.6	CAA ATG TAT GCY GGY TGG GCT GGT	+	5611–5634	1,201	This study
RCVr6.8	GYG CMG ACG AGT AAT TGT TTA GCG AC	−	6786–6811		This study
5g
RCVf6.5	CAT CTA YGS TGT TGC AAA TGG	+	6498–6518	954	This study
RCV end	dT(29)-ATA ATT TAC TCT AAA TTA TAA ACC AAT TAA ATT AAT TAA C	−	7383–7422 [+poly(A)]		This study
Additional primers
RCVr1.2	TGA GCT TSC CAG CDC CYT TCA TG	−	1203–1225		This study
RCVf0.8	AAT GCT GTT GCT GTG GAY ACA AC	+	805–827		This study
RCVf4.2	AAT TCD GGY AAR GCW CTC CAC CAT GT	+	4222–4247		This study

DNA library preparation and sequencing.

DNA libraries were prepared using 0.7 ng of pooled amplicon DNA and the Nextera XT DNA sample preparation kit (Illumina) according to the manufacturer's instructions. Nextera XT DNA libraries were indexed and sequenced using the Illumina MiSeq platform (300 cycle v2 kit). Sequence read quality was assessed and reads were trimmed as described previously (5). To create a consensus sequence for each sample, reads were mapped to the RCV-A1 reference sequence [EU871528.1 AUS/MIC-07(1-4)/2007.37] or the RHDV reference sequence (M67473.1 DEU/FRG/1988.50) using the Geneious package v8.1.5 (62). Primer sequences were trimmed from the terminal ends of the genome.

Compilation of sequence data sets.

The complete genome sequences of 11 RHDV isolates and 44 RCV-A1 isolates (43 from Australia and 1 from New Zealand) sequenced in this study were aligned with all RCV (Australian RCV-A1 and European RCV-E) and RHDV (classic RHDV and RHDV2) sequences available in GenBank (http://www.ncbi.nlm.nih.gov/GenBank/index.html). Genome sequences were aligned using MUSCLE as available in the Geneious package with default settings. From this alignment, three RHDV/RCV data sets were constructed, comprising (i) complete genomes (coding) (n = 146 sequences, 7,351 nucleotides [nt]), (ii) nonstructural protein genes (n = 148, 5,283 nt), and (iii) the VP60 gene (n = 338, 1,740 nt). Note that the VP60 data set is significantly larger than the genome and nonstructural data sets, as for many viruses only the VP60 gene was sequenced, including RCV-E. Separate RCV-A1-only data sets were also created for the VP60 gene (n = 61, 1,740 nt), nonstructural genes (n = 45, 5,274 nt), and the full genome (n = 45, 7,342 nt).

Phylogenetic analysis.

The jmodelTest program v2 (63) was used to determine the best-fit model of nucleotide substitution, which was found to be GTR+I+Γ₄ in all cases. Maximum-likelihood (ML) trees were then estimated for the nonstructural genes and VP60 gene data sets using PHYML v3.1 (64) as available in Geneious (62). All phylogenies were inferred using a combination of nearest-neighbor interchange (NNI) and subtree pruning and regrafting (SPR) branch swapping, with branch support estimated using 1,000 bootstrap replicates. All trees were rooted using a European brown hare syndrome virus (EBHSV) sequence (accession number KC832839), which is known to be more divergent. Note that after initial analyses using the complete data sets, condensed VP60 (n = 112) and nonstructural gene (n = 94) phylogenies were produced, with reduced numbers of classic RHDV sequences, to minimize tree size. This did not affect the tree topology. Phylogenies containing the complete data sets are available from the authors upon request.

For both the complete genome and VP60 gene data sets, those amino acid substitutions that distinguished nonpathogenic (RCV-A1 and RCV-E) from pathogenic (classic RHDV and RHDV2) viruses were determined manually. Such “distinguishing sites” were defined as those where the benign viruses universally had an amino acid different from that of pathogenic viruses. Because full genome sequences have not been published for the RCV-E samples, only the VP60 sequences of these viruses were included.

Recombination detection.

The full genome alignment of all RCV and RHDV sequences was screened for recombination using the RDP, GENECONV, and MaxChi methods and BootScan (65–68) available within the Recombination Detection Program, version 4 (RDP4) (69). Only sequences with significant evidence (P < 0.05) of recombination detected by at least two methods and confirmed by phylogenetic analysis were deemed recombinant, and all recombinants were visualized using SimPlot (70). All recombinants detected here were removed from the full genome data set to prevent interference in subsequent analyses.

Analysis of selection pressures.

Site-specific selection pressures in RCV-A1 were analyzed using a variety of methods available through the Datamonkey web server of the HyPhy package v2.4 (http://www.datamonkey.org/), all of which estimated the ratio of nonsynonymous to synonymous substitutions per site (d_N/d_S ratio): fixed-effects likelihood (FEL), internal fixed-effects likelihood (iFEL), fast, unconstrained Bayesian approximation (FUBAR), single-likelihood ancestor counting (SLAC), and random-effects likelihood (REL). Those sites with P values of <0.05, a Bayes factor of >100, and/or posterior probability values of >0.95 were considered to provide significant evidence of positive selection under these approaches. In addition, we computed the d_N/d_S (ω) ratio of external (ω_e) versus internal (ω_i) branches for the RCV-A1 full genome, nonstructural gene, and VP60 gene data sets using the Codeml program available in the PAML package, v4.8 (71).

Evolutionary dynamics and times to common ancestry in RCV-A1.

Sampling dates (day/month/year) were assigned to all sequences. To initially explore the extent of clock-like structure in the RCV-A1 sequence data, we plotted root-to-tip genetic distances from the ML tree against the time of sampling using the Path-O-Gen v1.4 program (http://tree.bio.ed.ac.uk/software). As this analysis revealed sufficient temporal structure (see Results), we next estimated evolutionary rates, as the number of substitutions per site per year (subs/site/year), using the Bayesian Markov chain Monte Carlo (MCMC) method available in the BEAST package v1.8 (72). This also allowed us to estimate the time to most recent common ancestor (TMRCA) of the RCV-A1 data. Based on initial analyses using the GMRF Bayesian skyride model of population growth (available from the authors on request), it was determined that the constant population size coalescent model was the most appropriate description of the data. In addition, marginal-likelihood estimation (MLE) using path sampling/stepping-stone sampling as available in BEAST revealed that the strict clock model and the relaxed uncorrelated lognormal deviation clock model equally fit these data (log MLE values differed by <0.5). Accordingly, the simpler strict clock model was selected. Each BEAST analysis was run for 50,000,000 to 100,000,000 generations until convergence was achieved, and at least two independent runs were performed for each set of priors. A maximum clade credibility (MCC) tree with mean node heights and Bayesian posterior probability values indicating the degree of support for each node was created using the TreeAnnotator program from the posterior set of trees. To further assess the robustness of the BEAST analysis, we performed a Bayesian randomization test in which each sequence was assigned a random date (73) and the BEAST analysis repeated. This date randomization test was repeated nine times and compared to the BEAST results with the correctly dated sequences described above. Finally, both the substitution rate and the time scale of RCV-A1 evolution were estimated using the Least Squares Dating (LSD) software v0.2 (74), using an unrooted tree and default settings.

Accession number(s).

The sequences of the genomes sequenced in this study have been submitted to GenBank under accession numbers KX357653 to KX357707 (Table 1).

RESULTS

Evolutionary history of rabbit lagoviruses.

We first inferred maximum-likelihood phylogenetic trees using the nonstructural gene (n = 148) and VP60 gene (n = 338) data sets, incorporating sequence data newly generated here together with those published previously (phylogenies with data sets of reduced size are shown in Fig. 3). The newly sequenced RHDV samples clustered with existing Australian field strains in both the nonstructural and VP60 gene trees. Similarly, the RCV-A1 sequences generated here clustered with those described previously (9), and the full-set of RCV-A1 sequences formed a monophyletic group, although this was only weakly supported in the VP60 phylogeny, and reconstruction of this evolutionary history is complicated by recombination (see below). Although RCV-A1 sequences generally grouped according to sampling location, at certain locations viruses from multiple clades were detected, either contemporaneously (CAT and MIC) or in subsequent years (Gudg, GUN, and OC/OAK). Hence, there has clearly been mixing among RCV-A1 populations, and in some locations multiple strains have circulated either alternatingly or simultaneously.

FIG 3.

Phylogenetic trees of RHDV and RCV nonstructural genes (n = 94) and the capsid gene, VP60 (n = 112). Maximum-likelihood trees of the nonstructural genes (A) and the VP60 (capsid) gene (B) are shown. Samples newly sequenced in this study are shown in bold, and Australian RCV-A1 taxa are colored according to the state in which they were isolated. The accession numbers for previously published sequences are indicated in the taxon name. The European (RCV-E) sequences and additional previously published RCV-A1 sequences were included in the VP60 tree only, as the full genomes were not available. The putative recombinants detected in this study, Gudg-26, Gore-425A (from New Zealand) (RCV-A1), and MRCV (semipathogenic) are highlighted yellow and boxed in blue, red, and green boxes, respectively (although the Gore-425A breakpoints are not between the RdRp and capsid, these trees are still clearly incongruent). The phylogenies were rooted using an early European EBHSV strain (not shown), and the scale bar is proportional to the number of nucleotide substitutions per site. Bootstrap support values are indicated at the major nodes.

There are some striking topological differences between the VP60 and nonstructural genes trees. The VP60 phylogeny shows that RHDV, RHDV2, RCV-E/MRCV, and RCV-A1 form distinct groups, with the RCV-E and MRCV sequences clustering with RHDV and RHDV2, suggesting that they are more closely related to these pathogenic lagoviruses than the more divergent RCV-A1 (Fig. 3B). A very different phylogenetic pattern is seen in the nonstructural gene phylogeny, in which the recombinant RHDV2/RCV-A1-like European strains cluster within the RCV-A1 lineage (Fig. 3A). This suggests both a potentially large amount of unsampled lagovirus diversity in Europe and some recent mixing of viral populations between Europe and Australia. MRCV also clusters closely with RCV-A1 in the nonstructural gene tree, indicating that this semipathogenic virus has a recombinant history. Finally, the New Zealand RCV sequence clustered within the diversity of RCV-A1 sequences, indicating a mixing of RCV isolates (and likely rabbits) between New Zealand and Australia (Fig. 3A and B).

Evidence for recombination in RCV-A1.

We identified two recombinants among the RCV-A1 sequences described here, Gudg-26 and Gore-425A (NZ) (Table 3). Gudg-26 had a single probable breakpoint between the capsid and RdRp that likely reflects template switching between the full genome and subgenomic RNA during replication (Fig. 3). Gore-425A (from New Zealand) had two breakpoints within the nonstructural genes. There was strong phylogenetic support for recombination in both cases, with high bootstrap values (Fig. 3 [Gudg-26] and data not shown [Gore-425A]). In addition, there was some ambiguous signal for recombination in viruses WAU-1, MIC-07(1-4), BUR1-1, and CAT3-4, although this did not result in significantly incongruent phylogenetic trees. Additionally, our analysis suggested that the semipathogenic strain MRCV is a recombinant between an RCV-E-like strain (VP60 gene) and an RCV-A1-like strain (nonstructural genes) (Fig. 3), with phylogenetic incongruencies again receiving strong bootstrap support. Gudg-26, Gore-425A, and MRCV were removed from the selection, regression, and full-genome Bayesian analyses. Since breakpoints were detected in the nonstructural genes for Gore-425A, this sequence was also removed from the Bayesian analyses of the nonstructural genes.

TABLE 3.

Recombination within the RCV-A1 and MRCV genomes

Sample	Putative parental strain		Breakpoint (nt)		P valuea determined by:
	Major	Minor	Start	End	RDP	GENECONV	BootScan	MaxChi
Gudg-26	GUN-238b	OAK NT-12b	5316	UDc (7359)	2.1 × 10−61	3.4 × 10−46	4.6 × 10−60	4.0 × 10−20
Gore-425A (New Zealand)	LWR-1b	WAU-1b/unknown	3961	4925	5.0 × 10−9	7.3 × 10−1	2.0 × 10−4	2.6 × 10−2
MRCV	GUN-343b	Parkwoodd	5451	UD (7212)	6.2 × 10−1	NDe	3.5 × 10−10	1.3 × 10−10

^aSignificant P values (<0.05) are italicized.

^bRCV-A1.

^cUD, undetermined (i.e., no breakpoint found).

^dRHDV.

^eND, recombination was not detected using this method.

Natural selection on RCV-A1.

Six codon positions were detected as subject to putative positive selection in the RCV-A1 genome by one or more methods: sites 22 (p16 protein; normalized d_N/d_S, 3.73), 56 (p16; normalized d_N/d_S, 1.72), 1385 (RdRp; normalized d_N/d_S, 1.95), 1836 (VP60; normalized d_N/d_S, 2.73), 2067 (VP60; normalized d_N/d_S, 2.82 to 4.01), and 2084 (VP60; normalized d_N/d_S, 3.93 to 10.52) [according to EU871528.1 MIC-07(1-4) numbering]. Overall, the evidence of positive selection was relatively weak and was most convincing at sites 2067 and 2084, with iFEL P values of <0.006 and <0.0006, respectively. Notably, site 2084, located in the P2 domain of the capsid, was the only site found to be under positive selection by three different methods. No selected sites were detected within ORF 2.

We also compared d_N/d_S (ω) values on external (ω_e) versus internal (ω_i) branches of the RCV-A1 phylogeny. This revealed a higher d_N/d_S ratio on external than on internal branches (Table 4), suggesting the presence of transient deleterious mutations or recent positive selection in our contemporary data set. Given the overall low d_N/d_S values in RC V-A1, we contend that the presence of transient deleterious mutations that are yet to be purged by purifying selection is the more feasible explanation, and this is commonly observed in data sets sampled over a short time frame (75). Such an excess of transient deleterious mutations is most pronounced in the capsid (VP60) gene, in which the d_N/d_S ratio for external branches is almost three times that for the internal branches (Table 4). The nonstructural gene data set had the lowest d_N/d_S ratio of external branches versus internal branches. Importantly, transient deleterious mutations may inflate estimates of evolutionary rate (and hence provide an underestimate of the TMRCA; see below) (5).

TABLE 4.

External versus internal d_N/d_S ratios for different RCV-A1 data sets

RCV data set	Mean ωa	ωeb	ωic	ωe/ωi ratio
Full genome	0.054	0.080	0.043	1.86
VP60 gene	0.053	0.087	0.032	2.72
Nonstructural genes	0.049	0.066	0.041	1.61

^aMean ω, average d_N/d_S over all branches of the phylogeny.

^bω_e, d_N/d_S ratio on external branches of the phylogeny.

^cω_i, d_N/d_S ratio on internal branches of the phylogeny.

Amino acid substitutions associated with changes in virulence.

All available RHDV, RHDV2, and RCV-A1 full-genome protein sequences were aligned along with that from the semipathogenic MRCV and the capsid protein sequences from RCV-E viruses. We then identified amino acid sites in which residues in all of the nonpathogenic (RCV) viruses differed from those found in RHDV/RHDV2 (Table 5). Four sites in the nonstructural proteins and five sites in the capsid protein distinguished pathogenic RHDV sequences from nonpathogenic RCV sequences and hence may in part determine virulence and/or tissue tropism. None of these sites were under positive selection. Of particular interest were those amino acid sites (positions 490, 2052, 2170, 2191, and 2218) where the semipathogenic MRCV (with liver tropism) differed from RCV but was congruent with RHDV and which may represent determinants for tissue tropism.

TABLE 5.

Amino acid changes that distinguish nonpathogenic RCV viruses from RHDV

^aAmino acid site number in ORF1 polyprotein according to the RHDV reference sequence (accession number M67473.1/DEU/FRG/1988.50) numbering.

^bFor the nonstructural proteins, the mature cleavage products are indicated in parentheses: p16, 2C (2C-like helicase), p29, and RdRp (RNA-dependent RNA polymerase). For VP60, the VP60-specific site numbering is provided in parentheses.

^cAmino acids located at each key site for benign Australian and New Zealand caliciviruses (RCV-A1), benign European caliciviruses (RCV-E), the Michigan semipathogenic lagovirus (MRCV), and pathogenic classic RHDV and RHDV2 strains (RHDV1/2), are indicated by 1-letter amino acid identifiers. RCV-E samples have only the VP60 sequence available and so were not included in the nonstructural alignment. Amino acids are shaded according to side chain pK_as and charge at physiological pH 7.4: green, positive side chain; yellow, polar uncharged side chain; gray, nonpolar side chain; blue and red, special cases (cysteine can be considered nonpolar or weakly polar, and glycine has no side chain).

Rate and time scale of RCV-A1 evolution.

To determine the strength of the clock-like signal in the RCV-A1 full-genome data set and hence the appropriateness of molecular clock dating, we performed a linear regression of root-to-tip distances on the ML phylogeny against the time of sampling. This revealed strongly clock-like evolution (correlation coefficient of 0.86) even though the sampling dates spanned a time frame of only 7.5 years (Fig. 4A). A similarly strong temporal structure was revealed using a Bayesian randomization test, in which the posterior distributions for estimates of nucleotide substitution rate and TMRCA in the “real” data set did not overlap those of the randomized data (results not shown).

FIG 4.

RCV-A1 temporal structure, rates of evolution, and time scale. (A) Temporal structure of the RCV-A1 full genome data set assessed by root-to-tip regression. The root-to-tip genetic distances (y axis) from the RCV-A1 full genome ML phylogeny were plotted against time (x axis), and a linear regression was conducted using Path-O-Gen v1.4. (B and C) Nucleotide substitution rate (B) and TMRCA (C) estimates for RCV-A1 sequences as measured using the Bayesian MCMC method for the complete genome data set (7,342 nt, n = 43), the nonstructural gene data set (5,274 nt, n = 44), and the VP60 gene data set (1,740 nt, n = 60). The y axis indicates the binned kernel density estimates of the posterior distribution.

Given the temporal structure in the data, we next estimated rates of evolutionary change (substitutions [subs]/site/year) and TMRCAs using a Bayesian MCMC approach. Regardless of the data set used, RCV-A1 was predicted to have a higher evolutionary rate than that previously reported for classical RHDV (Fig. 4B) (5). Specifically, mean rates of 5.3 × 10⁻³ to 5.8 × 10⁻³ subs/site/year were estimated, with 95% highest posterior density intervals (HPD) spanning 4.6 × 10⁻³ to 7.0 × 10⁻³ subs/site/year (Fig. 4B), compared to 2.47 × 10⁻³ to 3.08 × 10⁻³ (mean, 2.77 × 10⁻³) subs/site/year in RHDV (56). Linear regression analyses of the same data set using Path-O-Gen estimated a mean rate of 3.95 × 10⁻³ subs/site/year (95% confidence interval, 3.2 × 10⁻³ to 4.7 ×10⁻³).

Notably, the TMRCA varied considerably according to the data set used for Bayesian MCMC analyses (Fig. 4C). The TMRCA HPD intervals did not overlap between the nonstructural gene data set (1976.7 to 1984.8) and the capsid gene data set (1985.0 to 1993.2) (Fig. 4C), indicating that sample composition has a profound effect on rate and dating estimates, as shown with RHDV (5). Given that the d_N/d_S ratio on external branches for the VP60 data set is almost three times higher than that on internal branches (Table 4), it is likely that these data are strongly affected by an excess of transient deleterious mutations. Therefore, we determined that the nonstructural gene data set (with the lowest ω_e/ω_i ratio) was more likely to provide the most accurate estimates. Under these rates (5.3 × 10⁻³ subs/site/year), the RCV-A1 sequences sampled here were estimated to share a ancestor in the early 1980s (Fig. 4C). Similar to the Bayesian MCMC estimates, linear regression analyses predicted a common ancestor in the late 1970s (1975.4 [95% confidence interval, 1967 to 1981]), while the Least Squares Dating software estimated a common ancestor in ∼1982. Due to the high rates measured here and to be as conservative as possible, we also estimated the TMRCA of RCV-A1 based on the evolution rate proposed for RHDV (2.77 × 10⁻³ subs/site/year) (6) using Bayesian MCMC methods. Under this fixed rate, RCV-A1 sequences were estimated to have a common ancestor between 1953.3 and 1961.9 (95% HPD; mean of 1957.6).

DISCUSSION

RCV-A1 is a nonpathogenic lagovirus that causes enteric infections in the European rabbit populations of Australia and New Zealand (10). As infection with RCV-A1 affords partial cross-protection against fatal RHD, potentially limiting the efficacy of RHDV as a biocontrol agent (33), it is important to understand the evolutionary origins, history, and processes of this virus, particularly in comparison to RHDV.

While our phylogenetic analyses suggested that RCV-A1 sequences formed a monophyletic group, the phylogeographic separation of RCV-A1 is not as strong as originally thought (9), as there was clear evidence of the cocirculation of multiple strains either alternatingly or simultaneously in various locations. As the mechanical insect transmission of RCV-A1 over great distances is unlikely (9, 17), the past practice of using live rabbits for deliberate RHDV virus releases for rabbit control across Australia may in part explain this phylogeographic pattern. In particular, resident wild rabbits were captured, infected, and rereleased, and rabbit traps contaminated with RCV-A1 shed in feces may have been insufficiently decontaminated between use.

On a broader scale, phylogenetic analyses of all available RHDV and RCV isolates revealed that the benign viruses (RCV-A1 and RCV-E) do not form a single monophyletic group that is distinct from the pathogenic viruses (RHDV and RHDV2); in particular, the European nonpathogenic caliciviruses (RCV-E) cluster with classical RHDV (Fig. 3). Consequently, mutations associated with the evolution of increased virulence or attenuation, as well as changes in tissue tropism, would have occurred at least twice. From the phylogenetic analyses of currently available RCV and RHDV sequences alone it is impossible to determine whether high virulence evolved independently in the classic RHDV strains and in the new variant RHDV2 or whether high virulence evolved once in the common ancestor of these viruses and was then lost in RCV-E and MRCV. However, given the lack of evidence of a pathogenic variant prior to the 1980s, a scenario where both the classic RHDV and RHDV2 have separately evolved from a nonpathogenic common ancestor appears more likely. It is theoretically possible that virulent RHDV and RHDV2 emerged from a species jump and that we have not yet sampled the appropriate ancestral species. This theory is particularly apt in the case of RHDV2, which is not species specific and has been detected in multiple hare species (52, 76, 77). However, the notion that a species jump involved New World lagoviruses associated with Eastern cotton tail rabbits (Sylvilagus floridanus) (42) seems unlikely. Indeed, as all RHDV sequences fall within the diversity of viruses from O. cuniculus, the most parsimonious reconstruction from the current phylogeny is that RHDV was derived from an RCV-like virus within O. cuniculus. Consequently, for S. floridanus to be the true source of RHDV would mean that there must be (unsampled) S. floridanus lagovirus strains that fall within the diversity of O. cuniculus. This, in turn, would mean that benign O. cuniculus viruses spread to the Americas, infected S. floridanus, increased virulence, then returned to Europe in the guise of RHDV. Such an evolutionary scenario seems highly unlikely.

Since RCV-A1 differs from RHDV in pathogenicity, tissue tropism, and key mode of transmission (9, 10, 16, 37), it might be expected that these viruses would exhibit very different evolutionary dynamics. However, our analyses suggest that RCV-A1 and RHDV have evolved in very similar manners. Both viruses have evolved at broadly similar rates, with rates perhaps even elevated in RCV-A1, and both viruses show evidence of similar evolutionary pressures (dominated by purifying selection) despite the fact that they differ so greatly in virulence (5, 6, 56, 57, 78).

One of our most striking observations is that the rate of RCV-A1 evolution, at approximately 5 × 10⁻³ subs/site/year, is both relatively high for RNA viruses and likely even higher than that previously observed in the pathogenic RHDV (for which a mean rate of 2.77 × 10⁻³ subs/site/year was recently estimated [6]). An elevated rate for RCV-A1 is in keeping with recently published data indicating that, in vitro, the RCV-A1 polymerase has a higher replication rate than the RHDV polymerase (79), which in turn may result in more rapid mutation accumulation. However, as indicated by our branch-specific evolutionary analyses, the rates estimated here have also likely been artificially elevated to some extent by the presence of transient deleterious mutations that have yet to be purged by purifying selection. Indeed, it is now well established that evolutionary rates in RNA viruses are time dependent (75), such that the long-term evolutionary rate for RCV-A1 is likely to be lower than that estimated here.

Under the evolutionary rates inferred here, the common ancestor of all RCV-A1 identified so far in Australia occurred in the late 1970s to early 1980s and hence far more recently than previous estimates that suggested an origin in the 19th century (9). However, the earliest serological evidence of RCV-A1 in Australia was detected in rabbits from Victoria in 1972 (55% seroprevalence) (35), several years earlier than our molecular clock estimates for the date of RCV arrival in Australia. The earlier serological evidence of RCV-A1 acts as an independent calibration of RCV-A1 evolutionary history in Australia and gives credence to the theory that our short sampling time frame has resulted in increased rates and accordingly an underestimate of the timing of RCV-A1 emergence. Although the estimates are not wildly divergent, this emphasizes the need to exercise caution when using molecular clock dating. In light of this, we also estimated the TMRCA for RCV-A1 based on the lower evolutionary rate estimated for RHDV (6, 41, 43, 58); this placed the timing of the common ancestor of the RCV-A1 strains sampled here to the 1950s and remarkably close to the time when MYXV was introduced as a biocontrol agent.

Despite these uncertainties, most of the data presented here suggest that RCV was likely introduced into Australia after the introduction of MYXV as a biocontrol in 1950 (25). Indeed, it is theoretically possible that RCV-A1 was introduced into Australia with MYXV. Early preparations of MYXV brought from the Rockefeller Institute in the United States to Australia were material that had been maintained by regular rabbit passage for over 40 years (27). If RCV was present in any rabbit used during passage, then it could have been present in the material harvested for virus preparations (3). However, this would rely on the virus maintaining itself systemically for 8 to 10 days following MYXV inoculation, which is unlikely as RCV-A1 is not associated with a prolonged viremia (28). More likely is that RCV-A1 was brought into Australia within a live domestic rabbit. Although border protection in Australia has been relatively strict from 1950 onwards, the importation of domestic rabbits was allowed. Since RCV-A1 does not cause any overt disease and there would not have been any kind of screening test for RCV or RHDV at that time, there would have been no way of detecting such an incursion. Rabbitries provide ideal conditions for a fecal-orally transmitted virus like RCV-A1, providing a constant supply of susceptible animals, often at high densities. Indeed, the first nonpathogenic lagovirus was isolated from a domestic rabbit (30), and one of the Australian isolates (WAU-1) was also isolated from a commercial rabbit breeding facility (9, 33). A domestic-to-wild rabbit transmission of this virus is feasible given that rabbit feces from healthy domestic animals is not treated as hazardous material and is often used as garden fertilizer. In addition, there is evidence that rabbits or rabbit material may have moved between Australia and Western Europe in recent decades, particularly as RHDV2/RCV-A1-like recombinants were detected in Portugal (56), and have clearly recombined with a virus closely related to RCV-A1. Because the Portuguese recombinant RHDV2 nonstructural genes cluster within the RCV-A1 lineage (Fig. 3A), it seems most likely that RCV-A1 was brought into Europe rather than the opposite direction. Additionally, S. floridanus was bought into Europe in the late 1960s to restock hunting reserves (42), and it is not inconceivable that European rabbits could have been brought back into Europe from Australia for similar reasons.

Although our data suggest a relatively recent origin of RCV-A1 in Australia, it is also possible that RCV-A1 was present in Australia prior to the introduction of MYXV but that the drastic reduction in the size of rabbit populations due to MYXV caused a major population bottleneck in RCV-A1, in turn leading to the extinction of many of the deeper lineages. Accordingly, we have sampled only descendants from the single viral lineage that survived the bottleneck event. Indeed, it is striking that the time scale of RCV-A1 evolution estimated using the RHDV rate closely matches the introduction of MYXV as a biological control agent. The introduction of RHDV in the mid-1990s may have had a similar impact. Similar to the case with MYXV, the advent of RHDV resulted in massive population bottlenecks and local extinctions of Australian rabbit populations, and it is feasible that ancestral RCV-A1 lineages became extinct as a result. Wider sampling of Australian rabbits for RCV-A1, including those from South and Western Australia, would help to confirm the current distribution and to elucidate the origins of this virus.

Despite the lack of disease associated with RCV-A1, we found evidence of adaptive evolution at six amino acid sites, two of which were in the P2 domain of the capsid protein. In particular, one of the RCV-A1 codon sites found here to be under positive selection, site 2067, corresponds to RHDV site 2072, which was previously reported to be under positive selection (5). This residue forms part of an exposed loop (L1) on the RHDV capsid protein, which through cell- and animal-based experiments with synthetic peptides has been implicated in host cell interactions and the elicitation of neutralizing antibodies (14), and it is possible that the VP60 region containing this residue fulfils similar roles in RCV-A1. Reinfections of survivors are frequent and well characterized for RHDV, and while these do not cause renewed disease, they trigger a strong boost in antibody titers, in particular IgA (80). Similarly, RCV-A1 elicits a strong adaptive immune response (33), and there is evidence that it too causes reinfections (T. Strive, unpublished data). This suggests that protective immunity may be transient in both RHDV and RCV-A1, similar to what has been described for human norovirus (81–83). There is no evidence of true persistence or reactivation for either RHDV or RCV-A1, suggesting that the virus maintains itself within populations by constantly infecting and reinfecting susceptible animals, and variation at key sites representing potential neutralizing epitopes may therefore present a distinct advantage.

We also found evidence of recombination occurring in at least two RCV-A1 strains. Although recombination has been claimed to be an important evolutionary mechanism in caliciviruses, to date there is no evidence of recombination between classical RHDV and RCV-like viruses and hence no evidence that changes in virulence and tropism occurred via this process. In this context it is of interest to note that the semipathogenic American lagovirus with liver tropism (MRCV) has a recombinant history, with RCV-A1-like nonstructural protein genes and an RCV-E-like VP60 gene. However, it is unlikely that this recombination event enabled this virus to switch its tissue tropism, since both parental sequences are close relatives of benign enteric lagoviruses. In addition, recombinant RHDV2 strains possessing an RCV-A1-like set of nonstructural genes appear to be unchanged in phenotype (56). This lends support to the notion that point mutation, rather than recombination, is responsible for changes in virulence and tissue tropism in the case of these viruses.

We identified nine amino acid changes that distinguished virulent (RHDV) from avirulent (RCV-A1/E) viruses. Clearly, these changes need to be tested experimentally to determine if they play a role in virulence or changes in tissue tropism. Studies in murine noroviruses (also caliciviruses) have shown that a single amino acid change in the NS1/2 protein (equivalent to p16/p23) or in the P2 domain of the capsid protein can lead to changes in tissue tropism and virulence, respectively (84, 85). Similar changes in virulence or tissue tropism due to one or a few amino acid changes have also been shown for other viruses. For example, in the case of Venezuelan equine encephalitis virus, an amino acid mutation in the E2 envelope glycoprotein transformed an avirulent and replication-incompetent variant into a virulent, replication-competent variant (86). Regardless of whether the mutations identified here might have emerged following a host jump or changing environmental pressures or simply through antigenic drift, additional research into the role of these mutations in determining virulence and tissue tropism is warranted.