Spatiotemporal diffusion of the 2024 Oropouche outbreak in Cuba

Abstract

Oropouche virus (OROV) was reported in Cuba in May 2024 and rapidly spread throughout the country. Here, among 147 reverse-transcription polymerase chain reaction-positive cases identified from May to July 2024, we sequenced 39 whole genomes of OROV. Phylogenetic analysis revealed that all sequences formed a monophyletic cluster nested within the reassortant lineage, named OROV_{BR-2015–2025}, which has been circulating extensively in Brazil since 2023. Additional phylogeographic analyses demonstrated that the Cuban subclade probably originated from a single viral introduction from the Brazilian state of Acre in early February 2024, followed by cryptic circulation until its identification in May. The introduction probably occurred in the central region of the country, from which the virus spread and established secondary transmission hubs in the western and eastern regions. These findings underscore the capacity of OROV to spread well beyond South America, which was considered its endemic area of circulation.

Whole genomes of Oropouche virus were sequenced in Cuba from people infected during the 2024 outbreak, and phylogenetic analyses suggest that this Cuban subclade was introduced through a single introduction from Brazil in early 2024.

Main

Oropouche virus (OROV) was identified in Trinidad and Tobago during the 1950s and later in South and Central America¹. OROV generally causes a mild illness, although more severe manifestations such as meningitis, encephalitis, Guillain–Barré syndrome and, more recently, vertical transmission have also been reported^2,3. OROV has a single-stranded, negative-sense RNA genome divided into three segments, designated according to their size as L (large, 6.85 kb), M (medium, 4.36 kb) and S (small, 0.95 kb)¹. OROV is transmitted between mammals mainly through the bite of infected Culicoides paraensis midges, although Culex quinquefasciatus has been described as a potential secondary vector in urban settings^1,4,5.

Since 2024, OROV has spread beyond the Amazon region in Brazil, prompting the Pan American Health Organization (PAHO) to issue an alert regarding its circulation in the Americas⁶. On 27 May 2024, the Cuban Ministry of Public Health confirmed the detection of local OROV transmission in the country⁷. Acute febrile Illness (AFI) surveillance in Santiago de Cuba province identified a surge of cases that tested negative for dengue virus. Soon, the Arbovirus National Reference Laboratory at the Institute of Tropical Medicine ‘Pedro Kouri’ identified OROV as the cause of the outbreak. By 28 August, 506 confirmed cases had been reported across 99 out of 168 of the country’s municipalities⁷.

Recent studies demonstrated that the ongoing OROV outbreak in Brazil resulted from the sustained transmission and dissemination of a OROV reassortant lineage, here named OROV_{BR-2015–2025}, which probably emerged in the Amazonas state between 2010 and 2014⁸ and spread to non-endemic regions of the country across 2024⁹. Preliminary phylogenetic analyses on the S segment of Cuban samples of Santiago de Cuba and Cienfuegos provinces suggested that the Cuban sequences were closely related to 2023–2024 OROV sequences from Brazil¹⁰. To better understand the origin and temporality of the first-ever reported OROV introduction in Cuba, as well as the dynamics of its rapid spread throughout the country, we sequenced and analyzed full-length OROV genomes from human serum samples collected between 12 May and 9 July 2024, across 14 of the country’s 16 first-level administrative divisions.

In this period, samples from 217 suspected Oropouche cases were processed by real-time reverse-transcription polymerase chain reaction (RT–PCR)¹¹, and 147 (67.8%) tested positive for OROV mRNA. We selected 39 samples (26.5% of positive cases) for whole-genome sequencing based on Ct values (≤28.5) and geographical and temporal representativeness (Supplementary Table 1 and Extended Data Fig. 1a,b). Maximum likelihood (ML) phylogenetic analyses performed individually for each genomic segment revealed that all OROV sequences belong to the reassortant clade (OROV_BR-2015-2025) detected in Brazil⁸ (Extended Data Fig. 2).

Extended Data Fig. 1. Oropouche cases in Cuba and OROV genomes by epidemiological week and province.

a) The blue bars represent the number of Oropouche cases throughout the country, whereas the green bars represent the number of genomes generated in this study for each Epidemiological Week (EW). Therefore, OROV genomes were represented during all EW of the OROV Cuban outbreak. b) The blue bars represent the number of Oropouche cases by province until 9 July 2024, whereas the green bars represent the number of genomes generated in this study for each province.

Extended Data Fig. 2. ML phylogenetic analyses of the M, L, and S segments of Cuban OROV sequences.

Phylogenetic trees were constructed based on the M (a), L (b), and S (c) segments of 131 complete OROV genomes, including the 39 genomes generated in this study and 92 prototypical sequences representing other OROV lineages from different countries and years, as suggested by Naveca et al.⁸. The OROV_BR-2015-2025 clade is highlighted in light gray in each tree, and its statistical support (SH-aLRT) is annotated. Tree branches are drawn to genetic distance, as indicated by the scale bar at the bottom of each panel. Cuban and prototypical OROV sequences are represented in colors following the coding scheme shown in the left panel and their genotypic classification.

To resolve the phylogenetic relationship between Cuban and Brazilian OROV sequences with more resolution, we conducted a ML phylogenetic analysis on concatenated L, M and S segments of a dataset comprising all OROV sequences from Cuba and sequences representative of Amazonian subclades (AMACRO_I, AMACRO_II, AM_I, AM_II, AM_III and RR_I) described previously⁸. Analysis of the concatenated sequence datasets used in this study revealed no evidence of recombination. This analysis showed that Cuban sequences formed a strongly supported (approximate likelihood-ratio test (aLRT) 1.0) monophyletic cluster (OROV-CU) nested within the Brazilian subclades AMACRO_I and AMACRO_II (Fig. 1a). These subclades comprise sequences from Brazilian states distributed across northern (Amazonas, Acre and Roraima), southeastern (Rio de Janeiro and Espírito Santo) and southern (Paraná and Santa Catarina) country regions⁸. Notably, OROV sequences from cases imported from Cuba in Europe¹² also clustered with the OROV-CU subclade (Extended Data Fig. 3 and Supplementary Table 2).

Fig. 1. Spatiotemporal dynamics of OROV-CU clade.

a, ML phylogenetic tree based on concatenated OROV genomes (n = 263) from the OROV_{BR-2015–2025} clade, including Cuban sequences (n = 39) and Brazilian sequences (n = 224) representative of the Amazonian subclades (AMACRO_I, AMACRO_II, AM_I, AM_II, AM_III and RR_I) described previously⁸. Major subclades are annotated, and branch support values (SH-aLRT) are shown. The tree is drawn according to the genetic distance scale at the bottom the panel. b, Bayesian time-scaled MCC tree (n = 117) inferred with an asymmetric discrete phylogeographic model based on concatenated genomes of the OROV_{BR-2015–2025} subclades highlighted by a dashed line in the previous panel plus the earliest sequence sampled in Amazonas in 2015. This subset includes all sequences from the OROV-CU clade (n = 39), along with all available sequences of Brazilian subclades AMACRO_I (n = 34) and AMACRO_II (n = 43) sampled in northern (Amazonas, Acre and Roraima), southeastern (Rio de Janeiro and Espírito Santo) and southern (Paraná and Santa Catarina) states. Branches in the MCC tree are color-coded according to the inferred ancestral location of their basal nodes, as indicated by the legend on the right. The PP support of the OROV-CU clade is annotated. c, The map mirrors MCC tree colors and marks Cuba and relevant Brazilian states associated with the Cuban clade emergence. Diffusion rates supported by a BF >3 are indicated in the migration graph. The light–dark color gradient of lines represents the relative strength by which the diffusion rates are supported: substantial (BF 3.2–10), strong (BF 10–100) and very strong (BF >100).

Extended Data Fig. 3. International Dissemination of OROV-CU.

Maximum likelihood phylogenetic tree based on concatenated OROV genomes (n = 265) from the OROV_BR-2015-2024 clade, including Cuban OROV genomes (n = 39), and two OROV genomes sampled in Italy (n = 1) and the Netherlands (n = 1). Clusters are annotated with their respective names and statistical support values (aLRT). The tree is scaled according to the legend at the bottom of the panel.

To model the viral diffusion process between Brazil and Cuba, we performed a discrete Bayesian phylogeographic analysis of a dataset of concatenated segments comprising: the oldest sequence of OROV_{BR-2015–2025} clade detected in the Amazonas state in 2015, all OROV sequences belonging to AMACRO_I and AMACRO_II subclades, and the Cuban sequences of this study. This analysis confirms that Cuban sequences branched in a highly supported (posterior probability (PP) 1.0) OROV-CU subclade that most probably originated from a single introduction event from the Brazilian state of Acre (posterior state probability 1.0) (Fig. 1b). The Bayes factor (BF) tests showed significant support for a nonzero rate only between Acre and Cuba (BF 21.3), while connections between other Brazilian locations and Cuba show no significant support (BF <1.5) (Fig. 1c and Supplementary Table 3). The time of the most recent common ancestor (T_MRCA) of the OROV-CU subclade was estimated at 2024-02-10 (95% highest posterior density (HPD) 2024-01-04 to 2024-03-17).

We then use geographical coordinates of patient residential areas to reconstruct the fine-scale dispersion of the OROV-CU clade through a continuous spatial diffusion model with non-homogeneous dispersion rates (Fig. 2a). Our analyses indicate that the central provinces, Ciego de Ávila, Sancti Spíritus and Camagüey, served as the principal entry point and early epicenter of spread from early March through early July toward western and eastern regions. The virus dispersed westward via Mayabeque and Artemisa provinces, that seeded transmissions to Pinar del Río and Matanzas provinces in mid-May and to Ciudad de la Habana in late June. The eastern expansion originated through Holguín province, with subsequent dissemination to Santiago de Cuba and Guantánamo provinces in early May. Direct diffusion events from the central region to the province of Pinar del Río were also detected in early July.

Fig. 2. Spatiotemporal dynamics of the OROV-CU clade.

a, Spatiotemporal dissemination pattern of the OROV-CU clade (n = 39) across Cuban provinces estimated from the continuous diffusion phylogeography process. Lines indicate viral diffusion events, with colors corresponding to the time of occurrence. b, Frequency distribution of short (<5 km), medium (5–30 km), and long (>30 km) distance diffusion events of the OROV-CU clade calculated from the branches of 1,000 randomly selected trees from the posterior distribution of the continuous phylogeographic analysis. c, Weighted branch dispersal velocity of the OROV-CU clade through time (posterior median, solid line; 95% HPD, pale areas). d, Effective number of OROV-CU clade infections (Ne, y axis) over time estimated under the coalescent-based Bayesian Skyline model (posterior median solid line; 95% HPD, pale area).

Continuous phylogeographic reconstructions revealed that the median distance of viral diffusion events was 22 km (standard deviation 91 km; interquartile range 39 km) (Fig. 2b), remaining relatively constant throughout the study period. Our analysis also demonstrated that very-short-distance events (<2 km) constituted 8% of total viral movements, followed by short-distance (2–10 km, 22%), medium-distance (11–30 km, 33%) and long-distance events (>30 km, 37%). We estimated the OROV-CU median dispersion velocity within Cuba at 1.90 km per day (95% HPD 0.77–3.18 km per day), remaining relatively stable over the study period (Fig. 2c). During the same period interval, coalescent-based population modeling indicates that OROV-CU underwent an initial phase of slow growth until March, a rapid exponential growth from mid-March to mid-April, and a plateau from mid-April to the most recent coalescent event inferred from the tree in late April (Fig. 2d).

These findings confirm that the Cuban OROV outbreak is linked to the spread of the OROV_{BR-2015–2025} clade, which actively circulates in Brazil, and further reveal that all Cuban sequences were part of a monophyletic cluster (OROV-CU) nested within the Brazilian AMACRO_II subclade. Our phylogeographic analysis suggests that the Cuban outbreak resulted from a single virus introduction, probably from the Brazilian state of Acre.

The AMACRO_II subclade circulated in six Brazilian states from Amazonian (Acre and Rondônia) and non-Amazonian (Rio de Janeiro, Espírito Santo, Pernambuco and Santa Catarina) regions. Our findings indicate the virus was introduced to Cuba in early February 2024, concurrent with its arrival in non-Amazonian Brazilian states (mid-January to mid-March)⁹. This synchronicity makes a direct epidemiological link between these extra-Amazonian states and Cuba unlikely. Conversely, circulation of AMACRO_II subclade in Acre and Rondônia was traced back to August 2023⁸, providing sufficient time for local spread and subsequent export to Cuba in 2024. Moreover, Acre is a major departure hub for Brazil–Cuba flights (source: Cuban International Health Control Program), providing a plausible route for the AMACRO_II subclade’s introduction into Cuba. Despite the BF test strongly supporting an epidemiological link between Acre and Cuba, we acknowledge that geographic sampling bias, due to uneven data coverage in the western Amazon and across South America, could confound this result.

Our analysis suggests that the OROV-CU subclade probably arose in early February 2024, indicating a period of approximately 3 months of silent transmission before the virus was detected in Cuba in late May. Our demographic reconstruction supports that the OROV-CU subclade underwent an initial phase of slow expansion until mid-March. Moreover, despite the near-simultaneous detection of OROV transmission in eastern, central and western provinces by the end of May¹⁰, our phylogeographic results point to the central provinces as the site of initial introduction. Based on these findings, we propose that the OROV-CU subclade expanded slowly within the central provinces from early February to mid-March, explaining its initial undetected spread. The dissemination of the OROV-CU subclade to the eastern region coincided with a phase of rapid growth, which ultimately led to its detection once a large population size and nationwide spread were achieved.

A previous study of the OROV_BR-2015-2025 clade in Brazil estimated an average dispersion rate of 0.66–1.00 km per day across different Amazonian regions and also found that most (65%) dispersal events were over very short distances (<2 km), a pattern consistent with the typical flight range of Culicoides midges⁸. The OROV-CU subclade seems to have dispersed at a faster average velocity of 1.90 km per day, although the overlapping HPD intervals suggest that the rate estimates in Brazilian Amazon and Cuba are not statistically different. Nevertheless, the proportion of viral movements over distances greater than 10 km was much higher in Cuba (70%) than in the Brazilian Amazon region (30%). This indicates that the mobility of infected individuals was a more important driver of interprovincial viral spread in Cuba.

Our study has some limitations. First, geographic sampling bias can have a large impact on the accuracy of discrete and continuous phylogeographic reconstructions. Although we selected samples to ensure geographical and temporal representativeness, samples with high RT–PCR Ct values were excluded due to the low likelihood of sequencing success and only 26.5% of the cases identified in Cuba during the study period were sequenced. Second, we did not investigate negative dengue samples collected before the recognition of OROV in Cuba. The potential misdiagnosis of early OROV cases therefore represents a possible source of error in the temporal reconstruction. Finally, the primary vector of OROV in Cuba remains unconfirmed. While we have detected the virus in pools of Culex quinquefasciatus mosquitoes and of Ceratopogonidae spp. midges¹⁰, and the known vector Culicoides paraensis is present in the country¹³, the precise role of each species in the outbreak remains uncertain.

The autochthonous transmission of OROV in Cuba demonstrates the virus’s capacity to spread beyond South American areas traditionally considered endemic. This raises concerns about the potential involvement of unrecognized vectors. Given that environmental, climatic and demographic changes are global phenomena, it is not surprising that OROV could spread further across the Americas and beyond, being crucial to strengthen global efforts in enhancing surveillance, improving vector control strategies and advancing research into the factors driving virus transmission.

Methods

Ethics and inclusion

This study was designed through a partnership among the Institute of Tropical Medicine, Pedro Kouri (Cuba), the Instituto Leônidas e Maria Deane, Fiocruz, (Manaus, Brazil) and the Instituto Oswaldo Cruz, Fiocruz (Rio de Janeiro, Brazil), with the coordination and support of the Cuban Ministry of Health and the PAHO. Roles and responsibilities were agreed upon among collaborators ahead of the research, including having both a Cuban and a Brazilian researcher as a co-principal investigators.

The samples processed were obtained anonymously from excess material remaining after routine arbovirus testing, according to the case definition established by the Cuban Ministry of Health as part of the national clinical–epidemiological and laboratory surveillance to identify OROV transmission. Sex/gender information was not considered. Patient personal information was not used or published. Ethics approval was obtained from the Institute of Tropical Medicine, Pedro Kouri (CEI-IPK 54-24).

Previous work from this region (both from our team) was used to guide the design of this study, as well as connect our findings to similar research, and has been taken into account in the citations for this Brief Communication.

Case definition and studied samples

A clinically suspected Oropouche fever case is defined as a patient who resides in or comes from an endemic area or an area with active OROV transmission, or an area with an unusual increase of AFI cases that report fever or fever history (38 °C or more) and headache, accompanied by one or more symptoms as myalgia, arthralgia, chills, low back pain, photophobia or conjunctivitis, with illness resolution in 2–4 days in a favorable manner, with no focalization; or as a patient classified as AFI up to day 6 after symptom onset with a negative dengue IgM test. Suspected cases with a positive real-time RT–PCR result for OROV are considered as confirmed Oropouche cases. Between 12 May and 9 July 2024, samples from 217 suspected Oropouche cases were sent to the Arbovirus-NRL for molecular testing. Real-time RT–PCR confirmed 147 Oropouche cases; 39 (26.5%) were selected for genomic characterization based on cycle threshold (Ct) values (≤28.5), geographical representativeness and chronological distribution of sample collection by epidemiological week. Anonymized patient information data were used and are provided in Supplementary Table 1. Samples were collected from patients from 14 out of 15 provinces except for Las Tunas province and Isla de la Juventud municipality (both with no positive samples) (Cuban territory comprises 15 provinces (main island) plus the special municipality of Isla de la Juventud (a small island near the main island)). Among the 39 infected individuals, 33% (13/39) were male and 67% (26/39) were female, with a median age of 49 years (range 12–90 years). None of the 147 cases detected during the study period presented severe clinical manifestations.

RT–PCR testing for dengue, Mayaro and Oropouche viruses

Country-wide laboratory diagnosis of OROV infection was carried out by the Arbovirus-NRL. Acute serum samples collected 1 to 7 days after symptom onset were extracted using the QIAamp Viral RNA Mini Kit (Qiagen), according to the manufacturer’s instructions. RNA extracts were first tested for dengue by RT–PCR¹⁴. Dengue-virus-negative samples were tested using a multiplex real-time protocol designed to amplify OROV and Mayaro virus RNA.

OROV whole-genome sequencing and genome assembling

OROV genome amplification and sequencing libraries were prepared with COVIDSeq kit (Illumina) using OROV-specific primers on 2× 150 cycles paired-end runs on a MiSeq instrument (Illumina), as previously described⁸. Raw sequencing data were converted to FASTQ files on Illumina BaseSpace (https://basespace.illumina.com). FASTQ files were imported to Geneious Prime v2024.0.7, trimmed and assembled using a customized workflow. The GenBank sequences OL689332, OL689333 and OL689334 were templates for the S, M and L OROV segments, respectively.

OROV whole-genome genotyping

Complete OROV consensus sequences of L, M and S segments of all 39 newly generated genomes from Cuba were aligned with corresponding segments of a dataset of 92 sequences containing the prototype sequences of OROV (L: AF484424; M: AF441119; S: AY237111), Iquitos virus (L: KF697142; M: KF697143; S: KF697144), Perdões virus (KP691627; M: KP691628; S: KP691629) and Madre de Dios virus (L: KF697147; M: KF697145; S: KF697146), as well as OROV sequences representing the major M (M₁ and M₂), L (L₁ and L₂) and S (S₁, S₂ and S₃) clades⁸ (Supplementary Table 2). Sequences were aligned using MAFFT v7¹⁵ under the automatic selection of the appropriate strategy according to data size and submitted to phylogenetic reconstruction by ML using IQ-TREE v2.1.1¹⁶, under the GTR + G4 + I model as selected by the ModelFinder¹⁷ software. The tree branches’ support was evaluated with Shimodaira–Hasegawa aLRT (SH-aLRT)¹⁸, with 1,000 replicates. ML phylogenetic trees were visualized and edited with FigTree v1.4.4 (https://github.com/rambaut/figtree/releases).

Recombination analysis

Alignments of concatenated S, M and L genomic segments were constructed for different OROV_{BR-2015–2025} datasets and screened for potential recombination using the Recombination Detection Program (RDP) version 5¹⁹. The recombination breakpoints and potential parental sequences of recombinants were analyzed using nine different detection algorithms (RDP, Geneconv, Bootscan, Maxchi, Chimaera, SiSscan, PhylPro, LARD and 3Seq) implemented in the RDP5 package, with default settings. Only recombination events detected by at least three different methods and P values less than 0.05 were considered reliable.

OROV lineage identification

To identify the OROV_BR-2015-2025 lineage(s) circulating in Cuba, we generated a concatenated alignment of genomic segments L, M, and S, comprising all 39 newly generated genomes from Cuba along with a subset (n = 224) of sequences representative of the Amazonian subclades (AMARO_I, AMACRO_II, AM_I, AM_II, AM_III and RR_I) previously described⁸ (Supplementary Table 2). The resulting dataset (n = 263) was aligned and submitted to ML phylogenetic reconstruction as described above, under the GTR + G4 + I model as selected by the ModelFinder¹⁷ software. The tree branches’ support was evaluated with SH-aLRT¹⁸, with 1,000 replicates, and the ML phylogenetic tree was visualized and edited with FigTree v1.4.4 (https://github.com/rambaut/figtree/releases).

Discrete Bayesian phylogeographic analyses

To infer the spatiotemporal dynamics of the OROV Cuban epidemic, we generated a concatenated alignment of genomic segments L, M and S comprising: (1) all 39 newly generated genomes from Cuba; (2) all available sequences of the closely related Brazilian subclades AMACRO_I (n = 34) and AMACRO_II (n = 43) in which Cuban sequences were nested; and (3) the earliest OROV_{BR-2015–2025} sequence sampled in the Amazonas state in 2015 (Supplementary Table 2). The resulting dataset (n = 117) was aligned and submitted to ML phylogenetic reconstruction as described above. The temporal structure of the ML tree was estimated by performing a root-to-tip linear regression with TempEst v1.5.3²⁰, and the significance of the correlation between the collection date and genetic distance was evaluated using a Pearson correlation test, with corrections applied for multiple comparisons. The time-scaled phylogeographic tree was reconstructed using the Bayesian Markov chain Monte Carlo (MCMC) approach, implemented in the software BEAST 1.10²¹ with BEAGLE library v4²², to improve computational time. Bayesian trees were reconstructed using the GTR + G4 + I nucleotide substitution model, the nonparametric Bayesian Skyline coalescent demographic model²³ and a relaxed molecular clock model with a continuous-time Markov chain rate reference prior²⁴. The spatial diffusion from Brazil to Cuba was modeled using an asymmetric discrete phylogeographic model with a continuous-time Markov chain rate reference prior²⁴. The most relevant migration pathways between locations were identified by applying the Bayesian stochastic search variable selection approach²⁵, and significant nonzero rates were summarized using the cross-platform SPREAD application²⁶. We consider rates yielding a BF >3 as well-supported diffusion rates. In this analysis, Cuban sequences were assigned to the national level, while Brazilian sequences were classified at the subnational state level. MCMC simulations were run sufficiently long to ensure convergence (effective sample size >200) in all parameters as assessed in TRACER v1.7²⁷. The maximum clade credibility (MCC) trees were summarized with TreeAnnotator v1.10 and visualized using ggtree v3.2.1 R packages²⁸.

Continuous Bayesian phylogeographic analyses

To investigate the OROV dissemination dynamics within Cuba, a continuous phylogeographic analysis was conducted with a concatenated dataset consisting exclusively of the Cuban sequences. The viral diffusion process was reconstructed using software BEAST 1.10 as described above, with a relaxed random walk phylogeographic model with a Cauchy distribution²⁹, and an informative prior on root height inferred from the previous discrete phylogeographic analysis (normal distribution, mean 0.30, standard deviation 0.001). We assigned latitude and longitude coordinates to each sequence based on the patient’s place of residence (municipality), adding a slight jitter (0.01 to both coordinates) to tips sharing identical coordinates. Convergence of MCMC chains was accessed as described above. The SPREAD v1.0.7 software³⁰ was used to summarize migratory events and reconstruct the spatiotemporal diffusion patterns of the virus. These patterns were then visualized on maps generated using the ggplot2³¹ and naturalearth³² R packages. In addition, the branch dispersal velocity of the Cuban cluster was inferred from a subsample of 1,000 trees using the seraphim R package³³.

Statistical analyses

No statistical method was used to predetermine sample size. No data were excluded from the analyses. Statistical analyses of spatiotemporal and dissemination dynamics were conducted as described in the ‘Discrete Bayesian phylogeographic analyses’ and ‘Continuous Bayesian phylogeographic analyses’ sections.

Reporting summary

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

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41591-026-04411-9.

Author contributions

M.G.G. and F.G.N. contributed to study conception and study design. R.G., M.M.P., M.A., L.P., A.J.B., D.H., S.S., V.S., M.M., G.B., I.A., F.G.N. and M.G.G. contributed to laboratory work. R.G., M.M.P., M.A., L.P., A.J.B., D.H., S.S., J.R.d.A., C.P., M.R., L.F., L.G., V.S., M.M., G.B., I.A., S.R., V.K., F.G.N. and M.G.G. contributed to data analysis. R.G., V.S., G.B., I.A. and F.G.N. contributed to results visualization. G.B., I.A., F.G.N. and M.G.G. contributed to writing the original manuscript. R.G., M.M.P., M.A., L.P., A.J.B., D.H., S.S., J.R.d.A., C.P., M.R., L.F., J.L., L.G., V.S., M.M., G.B., I.A., S.R., V.K., F.G.N. and M.G.G. contributed to manuscript revision. M.G.G., F.G.N., L.F., L.G. and J.M. contributed to resources.

Peer review

Peer review information

Nature Medicine thanks Xiang Ji and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Lia Parkin, in collaboration with the Nature Medicine team.

Data availability

All the OROV genomes generated in this study were deposited at GISAID under accession numbers EPI_ISL_19611792–EPI_ISL_19611830 (10.55876/gis8.260330dk) (ref. ³⁴). Sample data can be found in Supplementary Table 1, and the sequence data used for phylogenetic analyses are provided in Supplementary Table 2. For further information, please contact the corresponding authors.

Competing interests

The authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s41591-026-04411-9.

Supplementary information

The online version contains supplementary material available at 10.1038/s41591-026-04411-9.