Rat hepatitis E virus (RHEV, species Rocahepevirus ratti), previously known as HEV species C genotype 1 (HEV-C1, species Orthohepevirus C), is the prototype of the genus Rocahepevirus in the family Hepeviridae of RNA viruses (Purdy et al., 2022). It is closely related to ferret hepatitis E virus (FrHEV/HEV-C2) of the same species. Since the first detection in Norway rats (Rattus norvegicus) from Germany (Johne et al., 2010), RHEV variants have been recorded in much of the world (Figure 1), with the Rattus species being the primary hosts. Recently, the population of Rocahepevirus has also expanded rapidly, with a growing number of novel members identified in various rodents from the families Muridae and Cricetidae (Reuter et al., 2020; Wang et al., 2020).

Within the subfamily Orthohepevirinae, rocahepeviruses are phylogenetically sister to though highly divergent from paslahepeviruses, which include the well-known HEV (species Paslahepevirus balayani, formerly Orthohepevirus A) that has been imposing a heavy burden on global public health. Four of the eight major genotypes of HEV (1, 2, 3 and 4) are usually responsible for human infection (Primadharsini et al., 2019; Reuter et al., 2020). Until recently, RHEV was demonstrated to be zoonotic like HEV-3 and -4 by a total of 23 patients from China (Sridhar et al., 2018; Sridhar et al., 2021; Sridhar et al., 2022), Canada (Andonov et al., 2019), Spain (Rivero-Juarez et al., 2022), and France (Rodriguez et al., 2023) (Table 1). Notably, immunocompetent individuals are also susceptible, and the clinical features vary greatly, spanning subclinical infection, acute or chronic hepatitis, extrahepatic manifestations, and even fatal outcomes. Moreover, likely spillover events have been observed in several other mammals (Table 1), particularly musk shrews (Suncus murinus), which might represent another viral reservoir (Wang et al., 2020).

RHEV has a positive-sense, single-stranded RNA genome of ∼6.9k nucleotides (nt). Between the 5′ 7-methylguanosine cap and the 3′ poly(A) tail, there are four open reading frames (ORFs) (Reuter et al., 2020; Wang et al., 2020) (Figure 2A). The longest ORF1 encodes a nonstructural polyprotein consisting of methyltransferase, papain-like cysteine protease, macro domain, RNA helicase, and RNA-dependent RNA polymerase (RdRp). Adjacent is ORF3 of the multifunctional phosphoprotein. This small ORF largely overlaps with ORF2 of the capsid protein. In addition to ORF4, which lies within ORF1, there are two putative ORFs (5 and 6) in the reference genome (NC_038504/GU345042); however, the two are absent in most of the complete genomes deposited in GenBank.

Given the ever-growing number of RHEV sequences in GenBank, we aimed to broaden the knowledge on the epidemiology and evolution of this emerging zoonotic pathogen. In this study, the available sequences of RHEV were first submitted to recombination detection. Following the phylogeographical analysis, the Bayesian coalescent approach was applied to a set of time-stamped sequences. In addition, the codon usage bias and dinucleotide composition of RHEV ORFs were measured.

Available complete and partial sequences of RHEV were downloaded from GenBank (as of August 2023). Multiple sequence alignments were created by using the MUSCLE program executed in MEGA X (Kumar et al., 2018) and then submitted to the RDP package (Takata et al., 2017). All built-in methods including RDP and BootScan were run to seek for possible recombination events. Sequence similarity comparison was visualized by SimPlot (Lole et al., 1999). To show phylogenetic incongruence, the sequences of the chimera and some representative isolates (information given as accession/country/host in Supplementary Figure S1) were segmented according to the suggested breakpoints. Each alignment was then submitted to MEGA X for reconstructing a Maximum Likelihood (ML) tree from 1,000 Bootstrap replicates.

To explain our genotype designation of RHEV, an ML tree was drawn with the partial genomic sequences (trimmed to region 554–6,090, position referred to GU345042) from 47 isolates (accession/country/host in Figure 3). Moreover, most isolates had genomic region 4159–4362 sequenced. Thus, to show the phylogeography and host range of RHEV, another ML tree was drawn based on the 204-nt region (though not sufficient for phylogeny reconstruction) using 131 representative isolates (accession/country/host in Supplementary Figure S2). The trees were generated under the best-fit nucleotide substitution model GTR + G + I determined by MODELTEST in MEGA X.

To date the divergence of RHEV, Bayesian analysis was conducted based on the genomic region 4193–4921 (3′ end of ORF1), given that many time-stamped isolates had the 729-nt region sequenced and a similar region was used for dating the divergence of HEV (Baha et al., 2019). The dataset was composed of 149 unique sequences (accession/country/year/host in Supplementary Figure S3) with collection dates and without ambiguous bases after data cleaning guided by TempEst (Rambaut et al., 2016). The subsets were compiled according to genotype designation. The times to the most recent common ancestor (tMRCAs) and the substitution rates were estimated by the Bayesian Markov chain Monte Carlo (MCMC) method in the BEAST v1.10.4 package (Suchard et al., 2018). Three clock models (strict, exponential, and lognormal) and two demographic models (constant size and exponential growth) were compared for confidence and convergence. The results from independent runs (each 10–25 million MCMC iterations with 10% burn-in) were pulled together to assure convergence (effective sample size > 200). Statistical uncertainty was reflected by the 95% highest probability density (HPD) intervals. The maximum clade credibility (MCC) tree was generated.

To measure the codon usage bias of the four RHEV ORFs, ENC (effective number of codons), GC_3S (frequency of G + C at synonymous 3rd codon position), and RSCU (relative synonymous codon usage) values were computed by using CodonW. ¹ The RSCU values of R. norvegicus genes were calculated based on the reference genome mRatBN7.2 in GenBank. ENC values vary between 20 (extreme bias as one synonymous codon is exclusively used) and 61 (no bias as synonymous codons are evenly used) (Wright, 1990). RSCU value > or < 1.00 stands for more or less frequent usage than expected, respectively. Moreover, to assess the potential effect of host (h) on virus (v) in codon usage, the dissimilarity index was transformed from Pearson Correlation Coefficient (based on the RSCU values of the 59 synonymous codons) using the equation D(v, h) = [1- R(v, h)]/2. The values were thus normalized into the range (0, 1). The higher the value, the larger the difference (Baha et al., 2019).

The dinucleotide composition of RHEV and R. norvegicus ORFs was measured by using DAMBE7 (Xia, 2018). The odds-ratio of observed frequency to expected frequency for dinucleotide XY was calculated as f _XY/f _X f _Y (f for nucleotide frequency). The upper and lower value boundaries are 1.23 and 0.78, respectively. An out-of-range value denotes that the dinucleotide is over- or under-represented (Cardon et al., 1994).

According to the phylogeography of RHEV (Supplementary Figure S2), while type b is restricted to East and Southeast Asia, type a has been detected in four continents (Figure 1). Notably, the Canadian patient was supposed to catch the infection during the stay in the Democratic Republic of Congo and Gabon (Andonov et al., 2019). Meanwhile, an a1-type sequence (MK935162) was sampled from R. rattus in Kenya (Onyuok et al., 2019). It is thus possible that RHEV is endemic in central Africa. Moreover, RHEV may be endemic in India and the USA, as suggested by the high seroprevalence rates of HEV antibodies in some Muridae species (Wang et al., 2020) [note that it is unknown whether the French patient was infected in Europe or India (Rodriguez et al., 2023)]. Due to the lack of sequence information in these regions, the origin of RHEV cannot be safely inferred from the current phylogeography. Nevertheless, there is little doubt that Southeast Asia is the cradle of the b lineage (Figure 3 and Supplementary Figure S2), though sampling bias is also present.

From the host information included in Figure 3 or Supplementary Figure S2, it can be derived that both a and b types can infect humans and shrews (listed in Table 1). Perhaps, RHEV is naturally infectious to most mammals and cross-species transmission is more dependent on the odds of close contact with the infected rats. Notably, all Rocahepevirus members except FrHEV are basically harbored by rodents. They can be split into two groups that infect Muridae and Cricetidae species, respectively, according to the ML phylogeny by Reuter et al. (2020). Given that FrHEV/HEV-C2 falls into the Muridae-infecting group, sharing with RHEV/HEV-C1 a common ancestor sister to HEV-C3 (tentative), it is highly possible that FrHEV represents an old host switch event from Muridae of Rodentia to Mustelidae of Carnivora. There is a special case that the isolate KU670940 from a common kestrel (Falco tinnunculus) is clustered with the rocahepeviruses from common voles (Microtus arvalis) (MK192405-9), but not the avihepeviruses (formerly Orthohepevirus B) from birds. It appears that hunting renders rocahepeviruses chances for host jumping from prey to predator.

In recombination analysis, three chimeric RHEV isolates (Figure 2B) were identified. It is not surprising that RHEV can undergo recombination to drive population variability, since several recombinants have been identified for HEV (Smith et al., 2020), one of which is even derived from a double event involving both intra- and inter-genotype recombination (Wang et al., 2010). In fact, it has been shown that the entire Hepeviridae family likely arose from an ancient recombination event occurring between plant and insect viruses. The breakpoint was located at the junction of the nonstructural and structural encoding regions, leading to an “alpha-like” ORF1 and a “Picorna-like” ORF2 (Kelly et al., 2016). Notably, the potential intergenotype recombination event of MH729810 also warns of rapid emergence of a distinct viral strain or species with possibly dangerous consequences.

According to Bayesian estimates (Table 2), the MRCA of the 149 RHEV isolates emerged just around 1956. Compared with HEV, RHEV has a much younger MRCA, which is even younger than the MRCA of HEV-1 (emerging around a century ago), the youngest one among the four human-infecting HEV genotypes (Purdy and Khudyakov, 2010; Baha et al., 2019). The average rate of RHEV was 1.06 × 10⁻² subs/site/year. This is quite high, although falling into the range of evolutionary rates documented for RNA viruses (10^–5 to 10^–1) (Sanjuan, 2012). It is similar to the rates of poliovirus 1 (PV1) (1.17 × 10⁻²) and PV2 (1.01 × 10⁻²), two members of the species Enterovirus C in the family Picornaviridae, but higher than the rates of most RNA viruses. In particular, it is ∼6-fold higher than the rate estimated for HEV (1.76 × 10⁻³) based on an 852-nt segment at the 3′ end of ORF1 under the same Bayesian models (Baha et al., 2019). It is not impossible that the RdRp of RHEV is more error-prone than that of HEV, since the two are highly divergent from each other. Although it is unknown whether the variance is associated with host difference, such high speed of evolution adds fuel to the zoonotic threat of RHEV.

In codon usage, RHEV is both similar to and different from R. norvegicus (Figure 4B). Such information might be useful in manipulating viral gene expression and designing attenuated viruses (Haas et al., 1996). It might be able to lower viral gene expression and virulence via deoptimizing the synonymous codons with those disfavored by both virus and host [e.g., “UUA” (L), “CGA” (R) and “GUA” (V)] and/or decreasing the number of the codons favored by both species [e.g., “CUG” (L) and “CAG” (Q)]. However, given the particular disfavor, increasing the number of “AGA” (R) and “GGA” (G), which are abundant codons in the host, might have a negative effect on viral viability rather than elevating viral gene expression. Notably, these might be extended to HEV, since the usage patterns of the mentioned codons are similar (Baha et al., 2019).

In mammalian genomes, “CG” and “TA” are markedly under-represented, which may result from DNA deamination following methylation and selection for increased mRNA stability, respectively (Simmonds et al., 2013). Such compositional abnormalities appear to be drawn on in cellular antiviral defense. In particular, zinc-finger antiviral protein (ZAP) has been identified as a powerful restriction factor active in “CG” and “UA” surveillance against non-self RNAs. Upon detection, ZAP can directly bind to the targeted sequences, leading to suppression of viral replication (Takata et al., 2017; Odon et al., 2019).

Unlike many other mammalian RNA viruses, including hepatitis A virus (HAV, species Hepatovirus A, family Picornaviridae) but not hepatitis C virus (HCV, species Hepacivirus hominis, family Flaviviridae) (Simmonds et al., 2013; Di Giallonardo et al., 2017), RHEV does not mirror hosts’ marked suppression of “CG” and “TA” (Figure 4D). Nevertheless, in RHEV ORFs, “UA” is particularly avoided at the (2, 3) codon positions (Figure 4B). Then, tRNA abundance is the likely factor accounting for the bias against the codons ending in “UA”, rather than ribonucleases such as RNAseL that can target “UA” for RNA degradation (Odon et al., 2019).

In general, mammalian hepeviruses are not particularly biased against “CG” and “UA”, nor are the closest relative viruses including rubella virus (RuBV, species Rubivirus rubellae, family Matonaviridae) and togaviruses (Rima and McFerran, 1997; Di Giallonardo et al., 2017). Perhaps, these viruses have developed an evasive/resistant strategy against “CG” and “UA” surveillance by ZAP and other antiviral factors, or even exploited the restriction to fine-tune the replication to maximize evolutionary fitness. Then, artificially increasing “CG” and/or “UA” frequencies, which is able to attenuate various RNA viruses (Odon et al., 2019), is not suitable for these viruses.

In summary, three RHEV sequences were identified as chimeric. Bayesian coalescent analysis with the time-stamped genomic sequences suggested that RHEV began to expand in the mid-20th century and was evolving at a very high rate. RHEV ORFs are rich in G/C and have additional bias independent of compositional constraints. In codon usage, RHEV is both similar to and different from the major host R. norvegicus. Moreover, RHEV does not mirror hosts’ marked suppression of “CG” and “TA”.