Rubella, which was first described in 18147, is an acute, highly contagious human infectious disease that is typically characterized by a rash, low-grade fever, adenopathy and conjunctivitis1. Research from the 1940s to 1960s revealed that the contraction of rubella (also called ‘German measles’) during the first trimester of pregnancy was directly associated with severe congenital birth defects, miscarriage and stillbirth8,9. Rubella virus (RuV), which is currently the only recognized member of the riboviriad family Matonaviridae (genus Rubivirus), is the aetiological agent of rubella10,11 and causes fetal pathology after transplacental transmission12. Extensive rubella epidemics have occurred worldwide due to the high airborne transmissibility of RuV (R0 = 3.5–7.8)13. Safe, efficacious, live-attenuated RuV vaccines, including the measles, mumps, rubella (MMR) vaccine, are now used worldwide and have successfully decreased the global incidence of rubella. However, around 100,000 cases of congenital rubella syndrome still occur annually1, and RuV can persist in immunologically privileged anatomical sites (for example, the eye) for years14. Furthermore, RuV infection in adults is generally underreported, as 30–50% of cases of adults with RuV infections are subclinical15. High-priority areas for rubella vaccination include the western Pacific, eastern Mediterranean and African regions, where RuV circulates widely and primarily infects young children16. The elimination of RuV is considered to be rapidly achievable because of the effectiveness of available vaccines and the lack of known animal reservoirs17,18.

Here we report the discovery of ruhugu virus (RuhV) and rustrela virus (RusV), which are relatives of RuV. RuhV was found in 10 out of 20 oral swabs from apparently healthy cyclops leaf-nosed bats (Hipposideridae: Hipposideros cyclops Temminck, 1853) in Kibale National Park, Uganda (Fig. 1). RusV was found in brain tissues of three acutely ill animals at a zoo in Germany, all of which succumbed to severe, acute neurological disease (Extended Data Table 2): a donkey (Equus asinus (Linnaeus, 1758)), a capybara (Hydrochoeris hydrochaeris Linnaeus, 1766) and a red-necked wallaby (Macropus rufogriseus Desmarest, 1817). RusV was subsequently detected in the brain tissues of 8 out of 16 yellow-necked field mice (Muridae: Apodemus flavicollis (Melchior, 1834)) on the zoo grounds and within 10 km of the zoo (Fig. 1 and Extended Data Table 1).

Fig. 1: Geographical locations of viruses and their hosts.
figure 1

a, Summary map of the estimated distribution of the cyclops leaf-nosed bat in Africa (red) and Uganda (blue box). b, Cyclops leaf-nosed bat in Kibale National Park, Uganda. Photograph credit: C. Johnson. c, Location at which the bat sample was collected and the ruhugu virus was discovered (Kibale National Park, Uganda, green star). d, Summary map of the estimated distribution of the yellow-necked field mouse in Eurasia (orange) and Germany (blue box). e, Yellow-necked field mouse in northeastern Germany. Photograph credit: U. M. Rosenfeld. f, Location of the zoo animals and discovery of RusV in Germany (southern Baltic Sea region, green star).

In the case of RuhV in Uganda, all bats were captured and sampled from five tree roosts (hollow cavities in trees) each of which contained between one and eight bats. Using molecular and metagenomic methods (Methods), RuhV RNA was detected in 5 out of 9 (55.6%) males and 5 out of 11 (45.5%) females in 4 out of 5 (80.0%) of the roosts (50% overall prevalence; 95% confidence interval, 29.9–70.1%). This high prevalence and frequency of positive roosts suggest that apparently healthy cyclops leaf-nosed bats are reservoir hosts, rather than incidental hosts, of RuhV. Cyclops leaf-nosed bats are small insectivorous bats that are primarily found in lowland rainforests from Senegal to Tanzania but are also found in coastal, montane and swamp forests as well as disturbed and agricultural landscapes19,20,21 (Fig. 1a), and are a host for Plasmodium cyclopsi, an apicomplexan ‘bat malaria’ parasite22,23. Whether RuhV can infect animals other than cyclops leaf-nosed bats remains unknown.

In the case of RusV in Germany, the donkey, capybara and red-necked wallaby were submitted for post-mortem evaluation and testing (Methods), which led to the identification of the virus (see below). Subsequent testing of rodents housed at the zoo and wild rodents on the zoo grounds and at two other locations within 10 km of the zoo revealed that 8 out of 16 (50%; 95% confidence interval 6.7–39.1%) yellow-necked field mice were positive for RusV RNA in brain tissue. Notably, the mice had no histological evidence of encephalitis (7 out of 8 mice investigated) and had only low concentrations of RusV RNA in peripheral organs (Extended Data Table 3). RusV RNA was not detected in any other small mammals collected simultaneously (n = 38; Extended Data Table 1). Yellow-necked field mice are omnivorous rodents that are native to parts of Europe and Asia, occupying habitats that range from mature forests to agricultural and peridomestic environments24 (Fig. 1d). They are a host of tick-borne encephalitis virus (Flaviviridae: Flavivirus)25, Dobrava virus (Hantaviridae: Orthohantavirus)26,27,28, Akhmeta virus (Poxviridae: Orthopoxvirus)29 and hepatitis E virus (Hepeviridae: Orthohepevirus)30. Routes of transmission of RuhV and RusV between reservoir hosts and to spill-over hosts (in the case of RusV) remain unknown, but the presence of the virus in oral swabs and faeces (Extended Data Table 3) suggests that contact with oral secretions and excreta could have a role.

Using molecular methods and in situ hybridization (Methods), we confirmed the presence of RusV in the brain tissues of all German zoo animals and in the liver of the donkey (Extended Data Table 2 and Extended Data Fig. 1). RusV RNA was detected within neuronal cell bodies and their processes in brain tissue sections of the donkey (Extended Data Fig. 1a), red-necked wallaby (Extended Data Fig. 1b) and capybara (Extended Data Fig. 1c) using in situ RNA hybridization. Histopathology revealed a nonsuppurative meningoencephalitis in all three animals, which was characterized by perivascular cuffing (Fig. 2a–c), meningeal infiltrates (Fig. 2d) and glial nodules (Fig. 2e). Neuronal necrosis and degeneration with satellitosis were detected in the brain stem of the donkey (Fig. 2f). Immune cells in the brain tissue consisted mainly of CD3-positive T lymphocytes, IBA-1-positive microglial cells and macrophages, and CD79a-immunoreactive B lymphocytes (Fig. 2g–l). In general, apoptosis was not a marked feature; only a few active-caspase-3-labelled cells were found to be distributed perivascularly and scattered within the grey and white matter (Fig. 2m, n). Multifocal perivascular red blood cells in the brain samples of the donkey and red-necked wallaby were positive for iron, as shown by Prussian Blue staining, which is indicative of intra-vital haemorrhages (Fig. 2o). The detection of viral RNA in samples from yellow-necked field mice collected between 2009 and 2020 and the absence of inflammation in the mice (Extended Data Fig. 1d, e) suggest that this broadly distributed rodent is the reservoir host of RusV.

Fig. 2: Histopathology and immune reaction of RusV in the brain of a capybara, red-necked wallaby and donkey.
figure 2

ac, Nonsuppurative meningoencephalitis with mononuclear perivascular cuffing in the brain of a capybara (a), red-necked wallaby (b) and donkey (c). d, Mononuclear meningeal infiltrates in the brain of a donkey. e, Glial nodules in the brain of a donkey. f, Neuronal necrosis (arrow) and degeneration with satellitosis (arrowhead) in the brain of a donkey. Haematoxylin and eosin was used. Scale bars, 20 μm (ac, e, f) and 50 μm (d). gl, Immunohistochemistry images of the immune reaction, in the perivascular tissue of the brain of a red-necked wallaby (gi) and in glial nodules of the brain of a donkey (jl). Numerous CD3-labelled T lymphocytes (g, j), IBA-1-positive microglial cells and macrophages (h, k) and CD79a-immunoreactive B lymphocytes (i, l) are shown. Immunohistochemistry was performed using AEC chromogen counterstained with Mayer’s haematoxylin. Scale bars, 20 μm. m, n, Apoptosis, indicated by few active-caspase-3-labelled cells (arrows) found in the perivascular tissue and scattered throughout the neuropil in the brain of a red-necked wallaby (m) and capybara (n). Immunohistochemistry was performed using AEC chromogen counterstained with Mayer’s haematoxylin. Scale bars, 20 μm. o, The Prussian Blue reaction highlights multiple iron deposits (arrows) within mononuclear cells that were found in perivascular tissue, mixed with accumulations of red blood cells, which is indicative of an intra-vital haemorrhage. Scale bar, 20 μm. Immunohistochemistry was performed on at least four slides per animal, yielding comparable results in all cases. In each run, positive control slides and a negative control for the primary antibodies were included. Evaluation and interpretation were performed by a board-certified pathologist (DiplECVP) with more than 13 years of experience.

The genome organizations of RuV, RuhV and RusV are identical, consisting of two large open-reading frames (ORFs), two untranslated regions at the 5′ and 3′ termini, and an intergenic region between the two ORFs (Fig. 3a). Across the non-structural and structural polyprotein-coding regions, RuhV is more similar to RuV than is RusV (Extended Data Table 4). Genetic similarity varies within the coding regions and is generally highest in a hyperconserved region within the Y domain of p1502,31,32 (Extended Data Fig. 2). RusV contains a markedly long intergenic region (366 nucleotides, compared with 46 nucleotides and 75 nucleotides in RuV and RuhV, respectively) and a correspondingly short C protein (205 amino acids, compared with 300 amino acids and 317 amino acids in RuV and RuhV, respectively; Extended Data Table 4). In addition, RuV and RuhV share a Gly-Gly-Gly amino acid sequence at the p150/p90 cleavage site, whereas RusV has a Gly-Gly-Ala amino acid sequence at this same site, which may impair cleavage in the case of RusV3.

Fig. 3: Evolutionary relationships among viruses.
figure 3

a, Comparative genome architecture of RuV, RuhV and RusV, showing five ORFs (coloured), two untranslated regions at the 5′ and 3′ termini (white) and an intergenic region (white) between the ORFs that encode the non-structural (nsPP) and structural (sPP) polyproteins. b, Maximum likelihood phylogenetic tree of RusV, RuhV and RuV genotypes 1A–1J and 2A–2C. Black silhouettes represent the natural hosts of each virus, and red silhouettes represent spill-over hosts in the case of RusV. Numbers beside nodes indicate bootstrap values (as a percentage; only values for major branches are shown); the scale bar indicates the number of amino acid substitutions per site.

RuhV (named for Ruteete subcounty, Uganda, and the Tooro word for insectivorous bat, obuhuguhugu) is an outgroup to all known RuV genotypes (Fig. 3b). RusV (named for its rubella virus-like genome and the Strelasund of the Baltic Sea in Germany) is a close outgroup to the clade comprising RuV and RuhV (Fig. 3b). This topology is consistent with the higher similarity of RuhV to RuV in each of the five mature polypeptides of the protein-coding viral genome (Extended Data Table 4 and Extended Data Fig. 2). Nucleotide sequences of RusV were 97.4–100% similar within the coding regions of the p90 and E1 genes sequenced in the samples from the donkey, capybara, red-necked wallaby and yellow-necked field mice in Germany (Extended Data Fig. 3).

The RuV E1 protein, a receptor-binding, class-II fusion protein5, contains an immune-reactive region (amino acid residue positions 202–283) with immunodominant T cell epitopes6 and four linear, neutralizing B cell epitopes (NT1–NT4)4 (Fig. 4a). The modelled tertiary and quaternary structures of trimeric E1 proteins of RuhV and RusV are homologous to the E1 protein of RuV33, and homology-based modelling of the quaternary structure of the E1 protein of RuhV predicts with high confidence that the E1 proteins of RuhV and RusV form homotrimers in the post-fusion state5 (Fig. 4b, c). One neutralizing epitope maps to amino acid positions 223–239 of the E1 protein at disulfide bond 8 (NT1)34. The mechanism of neutralization appears to involve blocking the trimerization of E1, which is necessary for virion fusion with the plasma membrane of the host cell5. Notably, only one amino acid residue (R237Q, near the C terminus) differs between the RuV and RuhV NT1 epitope (Fig. 4a), despite higher divergence at the amino acid level across E1 (Extended Data Fig. 3). By contrast, RusV differs from RuV at five amino acid residues within the same region (Fig. 4a). T cell epitopes are not well conserved in the capsid protein (Extended Data Table 5); however; the exposed putative linear epitopes of NT3 and NT4 in the E1 protein of RuhV and RusV are moderately conserved in comparison to RuV (Fig. 4 and Extended Data Table 5), suggesting that they should also be evaluated for cross-neutralization by anti-RuV antibodies.

Fig. 4: Comparisons of the E1 envelope glycoproteins of RuV, RuhV and RusV.
figure 4

a, Amino acid alignment and sequence logo of an immunoreactive region of E1 for RuhV, RusV and 13 RuV genotypes (GenBank accession numbers are included in parentheses). Lines indicate the locations of putative linear neutralizing B cell epitopes NT1–NT4. b, Homology-based model of the structure of the E1 homotrimer of RuhV in the post-fusion state, showing the receptor-binding site view (left) and profile view (right). Global model quality estimates (QMEAN) indicate a good model fit relative to the crystal structure of the E1 protein of RuV in the post-fusion form (Protein Data Bank biological assembly 4ADG_1). c, Homology-based model of the structure of the E1 homotrimer of RusV in the post-fusion state, as described above for RuhV. Key differences are seen in the modelled neutralizing epitopes NT3 and NT4 and in fusion loops 1 and 2 (FL1 and FL2). Residues of FL1 and FL2 of RuhV residues are highly similar to those of RuV, whereas FL2 residues of RusV differ from those of FL2 of RuV to a greater extent. The colour scale indicates the normalized QMEAN local score.

The fusion loops (FL1, residues 87–92; FL2, residues 130–136) in the E1 protein of RuhV are predicted to support the unusual metal ion complex that is necessary for E1-mediated RuV membrane fusion due to the presence in RuhV of amino acids N87 and D135 (homologous to RuV N88 and N136, respectively5; Fig. 4b). By contrast, FL2 of RusV is predicted to be less similar to RuV due to two amino acid residue replacements, P134A and T135A, the latter of which comprises a change from a polar to a non-polar residue (Fig. 4c). Across the RuV, RuhV and RusV genomes, regions of marked conservation and stabilizing selection are evident immediately upstream of the putative methyltransferase domain of p150, in the RdRp domain of p90, and proximal to the aforementioned NT1 epitope of E1 (Extended Data Fig. 2).

The similarity or near identity of certain RuV, RuhV and RusV B cell epitopes (Extended Data Table 5) suggests that existing serological assays for anti-rubella antibodies might detect RuhV, RusV and other as-yet-undescribed RuV-like viruses. Future studies that evaluate the performance of existing serological tests for RuV infection in animals would be useful, as would the development of new assays that can detect and differentiate among rubella-like viral infections in animals and humans. The implication that RuhV or RusV are zoonotic agents is currently speculative; however, bats and rodents possess biological attributes that predispose them to hosting many zoonotic viruses35,36,37, so this scenario should not be dismissed. The ability of RusV to infect both placental and marsupial mammals and to cause disease symptoms that resemble the severe encephalitic forms of rubella in humans38,39 reinforces such a precautionary stance.

The Global Measles and Rubella Strategic Plan of the World Health Organization (WHO) aims to control or eliminate rubella and congenital rubella syndrome in 5 out of 6 WHO regions by the end of 202040. Our discovery of relatives of RuV that infect asymptomatic bats and rodents suggests that rubella may have arisen as a zoonosis. Furthermore, the ability of RusV to infect mammals across wide taxonomic distances and to cause severe encephalitis in spill-over hosts raises concern about the potential for zoonotic transmission of RuhV, RusV or other RuV-like viruses. Despite these concerns, our findings will facilitate comparative studies of RuV that were previously not possible, including the potential development of animal models of rubella and congenital rubella syndrome.


Data reporting

No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.

Animal sampling and pathology

In Uganda, cyclops leaf-nosed bats were captured and released in Kibale National Park from June to July 2017. Kibale is a 795-km2 mid-altitude semideciduous forest park (0° 13′–0° 41″ N, 30° 19′–30° 31″ E)41 within the Albertine Rift, which is a region of exceptional biodiversity42 (Fig. 1c). Bats were caught in mist nets (Avinet) set in their flight path as they exited tree roosts at dusk and were kept in cloth bags until processing. Oral swabs were collected from each bat using sterile rayon-polyester-tipped swabs and preserved in 500 μl of TRI Reagent (Zymo Research). Swabs were frozen at −20 °C within 3 h of sample collection and transported on ice for storage at −80 °C before analysis. Animal collection and handling protocols were approved by the Uganda Wildlife Authority, the Uganda National Council for Science and Technology, and the University of Wisconsin-Madison Animal Care and Use Committee. Samples were shipped in accordance with international law and imported under PHS permit number 2017-07-103 issued by the US Centers for Disease Control and Prevention.

In Germany, a donkey, a capybara and a red-necked wallaby were submitted for necropsy from July 2018 to October 2019 after presenting with acute and severe neurological signs, including ataxia, convulsions, lethargy and unresponsiveness. All animals were housed at the same small zoo close to the Baltic Sea coast in northeast Germany (Fig. 1f). Standard diagnostic tests were negative for rabies virus, bornaviruses, West Nile virus, herpesviruses, Listeria, Salmonella and Toxoplasma. Formalin-fixed, paraffin-embedded (FFPE) brain tissues (cerebral cortex, cerebellum, brain stem and medulla oblongata) were cut at 3-μm thickness and stained with haematoxylin and eosin for examination using light microscopy. Conventional Prussian Blue staining was performed to demonstrate the presence of ferric iron, which indicates haemosiderin. Immunohistochemistry for immune cell markers was performed according to standardized procedures (Extended Data Table 6), and bright red intracytoplasmic chromogen labelling was produced with 3-amino-9-ethylcarbazole substrate (AEC, DAKO). Sections were counterstained with Mayer’s haematoxylin.

In situ hybridization for the detection of RusV RNA in brain tissue sections was performed with the RNAScope 2-5 HD Reagent Kit-Red (Advanced Cell Diagnostics) according to the manufacturer’s instructions. For hybridization, RNAScope probes were custom-designed against the RusV non-structural protein gene. The specificity of the probes was verified using a positive control probe against peptidylprolyl isomerase B (cyclophilin B) and a negative control probe against dihydrodipicolinate reductase (DapB). Histopathology and RNAScope interpretation were performed by a board-certified pathologist (DiplECVP).

Rodent management on the zoo grounds and hygiene measures for zoo staff were intensified after detection of a RusV infection in the deceased zoo animals. From September 2019 to February 2020, a total of 29 muroid rodents were collected from the grounds of the zoo (Extended Data Table 1). In addition, two brown rats (Rattus norvegicus) and three house mice (Mus musculus) housed at the zoo were sampled. Additional wild rodent samples were collected or retrieved from freezer archives from two trapping sites within 10 km of the zoo, where long-term research on rodent-borne pathogens is being conducted43. All wild-caught rodent species identifications were confirmed by cytochrome b DNA barcoding44. The zoo does not house bats and bats of the genus Hipposideros do not inhabit Germany. However, bats of the related and comparably speciose genus Rhinolophus do inhabit Germany and probably occur on or near the zoo grounds45.

All work with live animals and animal tissues was performed in compliance with all relevant ethical regulations.

Metagenomic, molecular and bioinformatic analyses

RNA was purified from bat oral swabs using the Direct-zol RNA MicroPrep kit (Zymo Research). RNA TruSeq libraries were then prepared, evaluated for quality, multiplexed and sequenced with NextSeq 500 v.2 chemistry using 2 × 150-bp cartridges (Illumina). RuhV was first identified using the VirusSeeker virus discovery pipeline46, after which deeper sequencing of two bat swab libraries was performed on a MiSeq (Illumina) sequencer using v.3 chemistry and 2 × 300-bp read lengths. The cyclops leaf-nosed bat genome was removed in silico by mapping reads to assembly PVLB01000001 using bbmap v37.7847 and discarding mapped reads. Non-viral reads were removed using FastQC v.0.11.5, bbmap v.37.78 and bbduk v.37.7847,48, and de novo assembly was then performed using metaSPAdes49. Reads were then mapped back to contigs for validation, related viruses were identified by DIAMOND using the BlastX algorithm49,50,51, and results were visualized using MEGAN v.652. Detailed analyses of contigs and reads were performed with CLC Genomics Workbench v.12 (QIAGEN).

Initially, red-necked wallaby and donkey tissues were processed using published methods for metagenomic pathogen detection53. In brief, tissues were first disrupted using the Covaris cryoPREP system (Covaris) and subsequently lysed in buffer AL (QIAGEN), followed by addition of TRIzol reagent (Life Technologies). After centrifugation, the aqueous phase was then transferred to RNeasy Mini kit columns (QIAGEN) and processed according to the manufacturer’s instructions, including on-column DNase treatment. Total RNAs from the cerebra of the donkey and the red-necked wallaby were used for library preparation53 and sequencing on an Ion S5 XL System with a 530 chip (Thermo Fisher Scientific). The RIEMS software pipeline54 was used for initial taxonomic assignment of reads.

After RusV RNA was confirmed in the donkey using the methods described above, deeper sequencing was performed on an Ion S5 XL System and a MiSeq (Illumina). The donkey genome was removed in silico by mapping reads to assembly ASM130575v1 using BWA55, and unmapped reads were filtered and retained. Read data quality trimming, adaptor removal and quality control were performed using the 454 software suite v.3.0 (Roche) and FastQC v.0.11.548. De novo assembly was performed using SPAdes v.3.12.056. RusV-specific contigs were then identified by DIAMOND using the BlastX algorithm51 followed by iterative mapping and assembly using the 454 software suite, SPAdes v.3.12.0 and Bowtie 2 v. for contig extension and verification. Results were visualized using Geneious (v.11.1.5, Biomatters). ORFs were identified by ORF Finder (implemented in Geneious). Conserved elements were identified by translated amino acid sequence alignment to RuV genomes using MUSCLE and subsequent annotation of p150, p90 and E1. The 5′ end of E2 was identified by the similar hydrophobicity and sequence pattern of the E2 signal peptide of RuV58 located at the C terminus of the capsid protein using ProtScale59 (window size 3; relative weight for window edges 100%; weight variation model linear). The 5′ terminus of the RusV genome was sequenced by rapid amplification of cDNA ends (RACE) using RNA from the donkey brain samples along with a 5′ RACE system v2 (Invitrogen) and specific primers.

FFPE brain tissues and peripheral organ samples from the donkey, capybara, red-necked wallaby, and wild-caught and zoo-housed rodents were assayed for RusV using an original one-step real-time quantitative reverse-transcription PCR (RT–qPCR). Total RNA from FFPE tissues was extracted using a combination of the Covaris truXTRAC FFPE total NA kit and the Agencourt RNAdvance Tissue Kit (Beckman Coulter). Nucleic acid extraction from unfixed rodent tissues was performed using the KingFisher 96 Flex Workstation (Thermo Fisher Scientific) and the NucleoMagVET kit (Macherey-Nagel) according to the manufacturer’s instructions. RT–qPCR was then performed using the SensiFAST Probe No-ROX One-Step kit (Bioline) with forward primer (1072–1091, 5′-CGAGCGTGTCTACAAGTTCA-3′), reverse primer (1219–1237, 5′-GACCATGATGTTGGCGAGG-3′) and 5′ probe (1161–1178, 5′-FAM-CCGAGGAGGACGCCCTGTGC-BHQ-1-3′) on a Bio-Rad CFX96 qPCR instrument (Bio-Rad). Primer and probe specificity were verified by BLASTn51 in silico analyses and Sanger sequencing of amplicons (Eurofins Genomics Germany), with the β-actin (Actb) gene used as an internal inhibition control. DNase digestion and RNA purification of nucleic acids of RusV-positive yellow-necked field mouse brain tissues (KS20/923, KS20/928, KS20/1296, KS20/1340, KS20/1341, KS20/1342, KS20/1343 and Mu09/1341) were performed using the Agencourt RNAdvance Tissue kit or RNeasy Mini kit RNA clean-up protocol (QIAGEN). Total RNAs from the capybara and mice were then used for cDNA synthesis and library preparation (200-bp fragments) and sequenced on a Ion S5 XL System with an Ion 540 chip60. RusV consensus sequences were determined by iterative mapping and assembly with the 454 software suite v.3.0 with reference to the RusV sequence derived from the donkey (GenBank MN552442).

Phylogenetic analyses and predictions of protein functional domains

To characterize relationships among RuhV, RusV and known RuV genotypes (Fig. 3b), coding sequences of non-structural and structural polyproteins were first concatenated and aligned using MAFFT v.7.388. A phylogenetic tree of aligned amino acid sequences was then inferred using IQ-TREE software v.1.6.1261, with automated model selection (JTTDCMut+F+R3) and 500,000 ultrafast bootstrap replicates62. Phylogenetic analyses of the envelope glycoprotein E1 and the helicase and RNA-directed RNA polymerase p90 (Extended Data Fig. 3a, b) were conducted as described above.

Prediction and annotation of the functional domain of proteins from RuhV and RusV were performed using the InterPro webserver63, and the confidence of E1 structural homology was estimated using Phyre233. Homology modelling of the quaternary structure of the post-fusion E1 homotrimer (Fig. 2c, d) was performed using the SWISS-MODEL workspace64 with model view by NGL65 and the residue colour corresponds to the local QMEAN score66, with 53 C-terminal residues of E1 (representing the stem and transmembrane segment of the E1 linear peptide) removed before homotrimer modelling5. Patterns of selection across the RuV, RuhV and RusV genomes were examined using SNAP 2.1.167,68.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.