Ecological factors and genetic features are associated with ecological generalism in pathogenic tick-borne viruses

Abstract

Emerging and re-emerging tick-borne viruses (TBVs) have caused numerous outbreaks recently, yet the factors driving their emergence and pathogenicity in humans remain poorly understood across viral taxa. A large-scale analysis integrating epidemiological and genomic data of pathogenic TBVs reveal that highly pathogenic TBVs (HPTBVs) tend to exhibit ecological generalism. Specifically, HPTBVs infect a broader range of tick species, animal hosts, and ecotypes, which likely enhance their transmission to humans. Bird migration is identified as the primary driver for geographic spread of HPTBVs through modeling and phylogeographic analyses, while livestock density and TBVs’ diversity they carry determine HPTBVs local persistence. HPTBVs experience stronger positive selection in genes related to host interactions, with specific mutations associated with lineage adaptation. Notably, the positively selected D170N mutation in severe fever with thrombocytopenia syndrome virus is predicted to be an enhanced binding affinity to human receptors, which correlate with increased infectivity. In this work, using TBV as a model system, we verify our hypothesis that ecological generalism is associated with increased pathogenicity and adaptability. These findings highlight the ecological and evolutionary mechanisms that drive the generalization of TBV and underscore their potential implications for public health risk.

This study finds that the most dangerous tick-borne viruses infect many tick and animal species, boosting their chances of reaching humans. Bird migration helps spread them worldwide, and genetic adaptation may increase their ability to infect people.

Introduction

Ticks are globally distributed, obligate, hematophagous ectoparasites, responsible for transmitting numerous tick-borne viruses (TBVs), including highly pathogenic ones like Crimean–Congo hemorrhagic fever virus (CCHFV), tick-borne encephalitis virus (TBEV), and severe fever with thrombocytopenia syndrome virus (SFTSV). These viruses pose significant public health risks and cause substantial economic costs. Advances in meta-transcriptomic sequencing have uncovered increasing TBV genetic diversity, highlighting spillover risks¹. To date, 870 TBVs have been documented, with 59 jumping to humans or animals, accounting for 75% vector-borne diseases². Remarkably, 13 TBVs have been newly detected in approximately the last decade: Alongshan tick virus (ALSV)³, Bourbon virus (BRBV)⁴, Guertu virus (GTV)⁵, Heartland virus (HRTV)⁶, Jingmen tick virus (JMTV)⁷, Haseki tick virus (KSIV)⁸, Tacheng tick virus 2 (TcTV-2)⁹, Tacheng tick virus 1 (TcTV-1)¹⁰, Songling virus (SGLV)¹¹, Beiji nairovirus (BJNV)¹², Oz virus (OZV)¹³, Yezo virus (YEZV)¹⁴, and Wetland virus (WELV)¹⁵. More importantly, they were identified to be related to human disease or with evidence of serologic exposure. Among them, JMTV was included in the list of “Proposed Pathogen X and Other Pathogens of Concern” by the World Health Organization (WHO) in 2024¹⁶. While its role as a human pathogen remains under investigation, its widespread distribution and zoonotic potential warrant continued surveillance and research.

TBVs have also been regarded as less efficiently transmissible compared to airborne viruses, largely due to the blood-feeding behavior of ticks, which require several days for a single meal. However, the recent surge in TBV spillovers, indicative of accelerated transmission, demands careful examination. This increase may be linked to climate change, which drives the overlap of wildlife habitats and leads to more frequent encounters between hosts, resulting in increased tick-host interactions, thereby enhancing tick virome diversity and elevating the likelihood of tick-borne pathogen emergence due to spatial changes of animal hosts^17,18. The geographic expansion of tick species has also played a crucial role in the emergence of tick-borne pathogens in new epidemic regions¹⁹, including the invasive tick species Haemaphysalis longicornis, which has contributed to the spread of SFTSV²⁰. Although previous studies suggest that the emergence of TBVs is related to ecological changes, the primary drivers remain obscure, and our knowledge of the underlying ecological mechanisms is fragmentary due to limited analyses and the challenges associated with the opportunistic sampling of TBVs.

Even in the presence of ecological drivers that facilitate initial spillover events, the long-term success of viruses in new host environments ultimately depends on their ability to adapt to new hosts and transmission cycles. Adaptive evolution enables the virus to overcome host immune defenses, establish efficient replication, and ensure effective transmission within the novel host population^21,24, and successful virus emergence events are often associated with a variety of host adaptations²². In this respect, it is notable that TBVs exhibit diverse phenotypes across various tick species and ecotypes, reflecting their genetic adaptability into new hosts and environments¹. However, the inherent genetic markers enabling TBV spillover and sustained transmission remain insufficiently characterized.

In this work, we first built a global database of epidemiological and genetic data on pathogenic tick-borne viruses (PTBVs) and evaluated their ecological characteristics related to pathogenicity. Additionally, we identified key ecological drivers of PTBV emergence and dissemination using multifaceted modeling approaches. Furthermore, we analyzed the evolutionary signature of TBVs that might be linked to their pathogenicity through phylogenetic and structural analysis.

Results

Global database of PTBVs

We retrieved 30,208 studies related to the currently reported 32 PTBVs using the search terms specified in the Methods (Supplementary Fig. 1), with 22,080 studies sourced from PubMed and 8128 from the Web of Science. Additionally, we integrated 18,745 sequence records from the NCBI database, 1409 RNA sequencing data sets from the SRA, and 607 self-sampled data sets¹ to screen for PTBVs. After eliminating duplicates and applying our inclusion criteria, 11,816 records were compiled for the PTBVs database, in which most reports were for TBEV (29.4%, 3471 records), SFTSV (29.2%, 3454 records) and CCHFV (13.2%, 1562 records), as well as mainly from ticks (42.5%, 5027 records) and humans (42.2%, 4984 records). This revealed a troubling gap in the surveillance of PTBVs, which primarily focused on clinically significant viruses while neglecting assessments in other animal hosts (Supplementary Fig. 2).

Geographic distribution of PTBVs

We identified 32 TBVs across 104 countries on six continents, excluding Antarctica. TBEV was reported in 58 countries, mainly in Western Europe, while CCHFV appeared in 63 countries, primarily in Southern Europe, the Middle East, and Africa (Figs. 1a, S2, S3). The newly emerging SFTSV has been detected mainly in 10 countries in East Asia, as well as in the Middle East and Western Europe. Notably, there have been only 40 reports of Tamdy virus (TAMV) to date, but it has been detected in 11 countries across Asia and Australia. JMTV had 384 reports across five continents, indicative of wide geographic distribution. Overall, the presence of PTBV-positive ticks or animals in proximity to the reported human cases emphasizing their local persistence within the ecological cycle of vectors, hosts, and humans (Fig. 1a, Supplementary Figs. 3–6).

Fig. 1. Geographic generalization of pathogenic tick-borne viruses (PTBVs).

a Geographic distribution of 32 PTBVs. b PTBVs species richness in ticks, animal hosts, and humans at the country level. The virus icons on the map indicate the locations of the first discoveries of virus records. Source data are provided as a Source Data file. c The timeline of the first discoveries and the first human case records for PTBVs.

PTBV species number was higher in vectors than in animal hosts or humans across 34 countries, significantly in Austria, France, the USA, Kazakhstan, Kenya, Russia, and Turkey, highlighting emergence risks in these regions. China reported the highest species number (20 PTBVs), followed by USA (10 PTBVs), and Australia and France (9 PTBVs each, Fig. 1b). Northeastern Asia has become a hotspot for emerging PTBVs, with recent emergence including ASLV (2017, Inner Mongolia, China), YZV (2021, Japan), SOLV (2021, Heilongjiang, China), BJNV (2021, Inner Mongolia, China), OZV (2023, Japan), and WELV (2024, Inner Mongolia/Heilongjiang, China) (Fig. 1c).

Ecological generalization correlates with pathogenicity

Among the 32 PTBVs, 15 species have been considered as classical TBVs based on well‑defined pathogenicity, including TBEV, Powassan virus (OWV), Omsk hemorrhagic fever virus (OHFV), Kyasanur forest disease virus (KFDV), CCHFV, SFTSV, Colorado tick fever virus (CTFV), HRTV, Dugbe virus (DUGV), Erve virus (ERVEV), Nairobi sheep disease virus (NSDV), Bhanja virus (BHAV), Thogoto virus (THOV), and Quaranfil virus (QRFV) (Supplementary Table 1). The remaining 17 species were classified into emerging TBVs. To identify ecological and epidemiological factors associated with viral pathogenicity, we further subdivided classical TBVs into a relatively higher‑pathogenic group (KFDV, OHFV, POWV, TBEV, CCHFV, SFTSV, HRTV) and a lower‑pathogenic group (CTFV, DUGV, ERVEV, NSDV, BHAV, THOV, QRFV) based on well‑established clinical evidence, including documented disease severity and case‑fatality rates in humans (Supplementary Table 1). Notably, all HPTBVs can be transmitted both transovarially and transstadially. We found that PTBVs capable of transovarial transmission can infect a greater number of tick species (median: 14 vs. 4, p < 0.05, Fig. 2a), while those capable of transstadial transmission infected a more diverse range of animal hosts (median: 18 vs. 3, p < 0.05, Fig. 2a). Hence, the transmission mode of PTBVs might determine their capability for ecological colonization.

Fig. 2. Host range and preference of pathogenic tick-borne viruses (PTBVs).

a Illustration of viral genome characteristics, ecotype distributions, and transmission mode of PTBVs. The Wilcoxon rank-sum test was two-sided. No adjustment for multiple comparisons was applied. b The prevalence, case numbers, and ecological generalism index of PTBVs in humans according to the meta-review analysis. The prevalence is presented as mean values ± Standard Error of the Mean (SEM). Source data are provided as a Source Data file. c The prevalence and richness of PTBVs in each tick genus. d The prevalence and richness of PTBVs in each animal order or class. The box plots display the distribution of the data. The central line represents the median. The box indicates the interquartile range (IQR), spanning from the 25th to the 75th percentile. The whiskers extend to the minimum and maximum values. Source data are provided as a Source Data file.

HPTBVs occupied more diverse ecotypes (proportion of occupied landcover type, 91.7% vs. 16.7%; Wilcoxon rank-sum test, p < 0.001, Fig. 2a) in addition to wider country-level distribution (country numbers, 9.5 vs. 4; Wilcoxon rank-sum test, p < 0.05, Fig. 2a). Under the random-effects model, HPTBVs demonstrated a higher minimum infection rate compared to LPTBVs in ticks by molecular detection (2.50% [2.02–3.03%] vs. 0.89% [0.13–2.22%], Q = 16.11, df = 1, p < 0.0001, Fig. 2b) and in animals by serology test (9.37% [9.24–9.51%] vs. 0.39% [0.27–0.53%], Q = 8.71, df = 1, p = 0.0032, Fig. 2d, while there was no significant difference observed in humans. For instance, TBEV infects 28 tick and 102 animal species, CCHFV 46 tick and 48 animal species, and SFTSV 17 tick and 45 animal species (Fig. 2c, d). These findings suggested that HPTBVs had greater opportunities to persist and thrive in ecological niches, resulting in a higher frequency of spillover events to humans. Thus, the Ecological Generalism Index (EGI) was proposed here to quantify the degree of ecological generalization by integrating all epidemiological indicators. The index classified HPTBVs and LPTBVs with a high accuracy of 0.94, with all HPTBVs showing higher EGI than LPTBVs, except OHFV (Fig. 2b, Supplementary Figs. 7, 8).

Ixodes persulcatus harbored the most PTBVs (13 species), followed by two globally invasive species, Haemaphysalis longicornis (11 PTBVs) and Rhipicephalus microplus (9 PTBVs) (Supplementary Figs. 9a, 10). The animal order Artiodactyla possessed the most diverse range of species (19 species; Supplementary Fig. 9b), including 14 PTBVs from Ovis aries and 11 from Bos taurus. Six PTBVs, including five HPTBVs (CCHFV, TBEV, SFTSV, POWV, and KFDV), have been detected in birds, with a pooled prevalence of 19.37% ([14.41–24.73%], Supplementary Fig. 10). Notably, CCHFV, TBEV and SFTSV could be detected from migratory birds such as Anser spp., Spatula querquedula, Uria aalge, Laridae sp., and Hirundinidae sp. (Supplementary Fig. 10), suggesting the role of birds for global expansion of HPTBVs.

Ecological factors driving the worldwide spread of HPTBVs

Bayesian phylogeographic generalized linear models were applied to further evaluate the ecological factors responsible for PTBVs global dissemination. We focused on four PTBVs (CCHFV, SFTSV, TBEV and JMTV), which exhibited a global distribution from the above meta-analysis and have sufficient sequences for analysis. Notably, the number of migratory bird flyways, which reflected the frequency of bird migrations across one region, was a significant predictor of TBEV dissemination in Far-Eastern regions, with positive log effect sizes of correlation (0.102, 95% highest posterior density (HPD) intervals: [0, 0.87], Fig. 3a) at destination locations. The bird flyway frequency at the original locations also contributed to CCHFV dispersal in the Europe and Africa regions with negative correlation (−0.26, HPD: [−2.5, 0], Fig. 3b), while human population density showed positive contribution (0.17, HPD: [0, 1.4], Fig. 3b). Additionally, mammalian diversity at the origin were identified as a negative factor influencing CCHFV transmission between Asian and Africa regions (−0.52, HPD: [−3.4, 0], Fig. 3c). The presence of flyways between regions exhibited a significant positive correlation with the global dissemination of JMTV (2.9, HPD: [1.7, 4.2], Fig. 3d). However, the average annual frequency of international flights, import and export volumes of major livestock species didn’t show significant associations with any PTBVs dissemination. Collectively, these results highlight the contribution of migratory flyways to the rapid spread of PTBVs, particularly inter-continent movements, although different effects varied on PTBVs.

Fig. 3. Ecological drivers for the long-distance dissemination of high pathogenic tick-borne viruses (HPTBVs).

a–d Correlation between predictors and viral geographic transition frequency from General Linear Model (GLM) analyses in BEAST. Posterior summaries of the product (log effect size) of the log constant through-time predictor coefficient and the predictor inclusion probability (pooled across time periods) for Tick-borne encephalitis virus (TBEV), Crimean-Congo hemorrhagic fever virus (CCHFV) and Jingmen tick virus (JMTV) lineages were presented. Points indicated the posterior mean, while ranges represent the 95% highest posterior density (HPD) intervals. Location-specific predictors were included as both origin (O) and destination.

Ecological drivers for the local establishment of PTBVs

Three modeling approaches were conducted for the occurrence of all human pathogenic TBVs to identify ecological factors responsible for the probability of PTBVs presence within each 10 × 10 km geographic region. Among three models evaluated, Boosted Regression Trees (BRT) model (best model accuracy: 0.94) outperformed both Random Forest (RF) model (0.93) and Support Vector Machine (SVM) model (0.90) based on the leave-out testing data set (Fig. 4a, Supplementary Fig. 11). As expected, the number of PTBVs species carried by ticks and animal hosts acted as the primary contributors to human PTBV occurrence in the same regions, with important scores of 10% and 9%, respectively (Fig. 4b). Following these, five host-related factors ranked among the top ten important contributors: livestock density including buffalo, pig and duck, mammal diversity, as well as human population (Fig. 4b). This highlights the importance of PTBV circulation in animal hosts for their persistence in enzootic localities, especially livestock, due to their close interactions with humans. Also of note was that the East Asian-Australasian flyway may be the main reason for the global distribution of PTBVs (Fig. 4c), as the hotspots for emerging TBVs are located in northeastern Asia.

Fig. 4. Ecological drivers for the occurrence and diversity of pathogenic tick-borne viruses (PTBVs).

a Model performance for the best model of Boosted Regression Trees (BRT), Random Forest (RF), and Support Vector Machine (SVM). CVE: cross-validation error. Source data are provided as a Source Data file. b–e Importance of eco-climate factors for the occurrence of PTBVs and high pathogenic tick-borne viruses (HPTBVs) from the best BRT model. f Partial effect of PTBVs diversity in animals for human PTBVs richness by General Additive Model (GAM); The solid line represents the estimated partial effect from GAM, and the shaded area indicates the 95% confidence interval. All statistical tests were two-sided, and no additional multiple comparison correction was applied. g Partial effect of PTBVs diversity in ticks for human PTBVs richness by GAM. h Interaction effect between PTBVs diversity in animals and ticks for human PTBVs richness by GAM.

The rankings of the top ten responsible factors for the presence of HPTBVs varied slightly, with increased importance attributed to duck density, human population, solar radiation, and maximum temperature (Fig. 4b, d), while the number of PTBVs species in ticks and animals, as well as mammal diversity, still had a significant association with the presence of HPTBVs (Fig. 4b). Across all environmental factors, solar radiation, urbanization and cultivated vegetation were significant contributors to both PTBVs and HPTBVs, potentially enhancing opportunities for tick-human interactions (Fig. 4d, e).

We were particularly focused on hotspot regions in Asia, where frequent and ongoing emergence has been observed, as illustrated in Fig. 1b. To identify the key drivers of PTBV species number in human populations within each 10 × 10 km geographic region, a generalized additive model (GAM) was utilized. Factors included in the model explained 97.4% of the variance in human PTBV species number, with two factors demonstrating significant contributions: PTBV species number in animals (92.3%, Fig. 4f), PTBV species number in ticks (1.9%, Fig. 4g), and their interaction within same geographic regions(5.8%, Fig. 4h). Together with the BRT model results, this analysis highlighted the capacity of PTBVs to maintain reservoir hosts, determined their spillover potential to humans (as evidenced by PTBVs presence and their species number in human populations).

Greater positive selection on HPTBVs than LPTBVs

To examine whether HPTBVs show greater adaptive evolution than LPTBVs, the numbers of nonsynonymous to synonymous substitutions per site (d_N/d_S) were compared for PTBVs of different pathogenicity. HPTBVs had significantly higher d_N/d_S ratios on internal tree branches (median: 0.080 vs. 0.051, p < 0.05, Fig. 5a), thereby removing the impact of transient deleterious mutations that inflate d_N/d_S on tree tips. This was particularly evident in host-interacting genes, such as glycoproteins in Bandavirus, Orthonairovirus, Mivirus, Thogotovirus, and Quaranjavirus, as well as membrane, envelope, and coat proteins in Orthoflavivirus and Jingmen virus group. This difference in selection pressures was further supported by an analysis of site-specific selection pressures using the FEL, FUBAR, and MEME methods, which revealed that three PTBVs—CCHFV, SFTSV, and JMTV— exhibited the most positively selected sites (Supplementary Table 2). Notably, all three viruses exhibited positive selection on cell-entry proteins, especially glycoprotein N (Gn) in SFTSV and CCHFV, crucial for virion attachment, and membrane proteins involved in viral-host membrane fusion. Specific amino acid residues were predominant in each geographically clustered lineage of CCHFV, highlighting lineage-specific adaptations (Fig. 5b, Supplementary Table 4). A similar pattern was observed in JMTV, particularly at glycoprotein codon 24 (Fig. 5d, Supplementary Table 5) and membrane protein codon 53 (Fig. 5e, Supplementary Table 6). In SFTSV, a high proportion of the 170 N mutation was identified in sub-lineages of genotypes F, B-2, and B-3, which exhibited greater infectivity and pathogenicity compared to other genotypes²³ (Figs. 5b, 6a, Supplementary Table 3). The 170 N mutation was prevalent in regions with high incidence rates, such as Henan, Hubei, and Anhui provinces in China, as well as South Korea (Fig. 6b). Structural analysis demonstrated that the 170 N variant formed more interactions with the human pro-low-density lipoprotein receptor-related protein 1 (LRP1), including two hydrogen bonds with shorter bond distances to LRP1 (Fig. 6b, d). In contrast, the 170D variant formed only one salt bridge/ionic interaction, which was more distant (Fig. 6b, c). This structural feature likely explains the enhanced infectivity and pathogenicity associated with the 170 N mutation. In addition, JMTV and SFTSV demonstrated positive selection on codons within the polymerase protein, which may be associated with the optimization of viral replication (Supplementary Table 2). Collectively, these results suggest that adaptation in these proteins has facilitated their ability to infect human hosts and persist following spillovers.

Fig. 5. Molecular adaptation favors the pathogenicity of high pathogenic tick-borne viruses (HPTBVs).

a The ratio of nonsynonymous and synonymous substitutions per site (d_N/d_S) on internal branches in cell-interaction genes and replicase genes among HPTBVs, low pathogenic tick-borne viruses (LPTBVs), and non-pathogenic tick-borne viruses (NPTBVs) at the species level. The Wilcoxon rank-sum test was two-sided. No adjustment for multiple comparisons was applied. The box plots display the distribution of the data. The central line represents the median. The box indicates the interquartile range (IQR), spanning from the 25th to the 75th percentile. The whiskers extend to the minimum and maximum values. b–e Phylogenetic trees and positively selected amino acid mutation of Crimean-Congo hemorrhagic fever virus (CCHFV) glycoprotein, Severe fever with thrombocytopenia syndrome virus (SFTSV) glycoprotein, Jingmen tick virus (JMTV) glycoprotein and membrane protein.

Fig. 6. Comparative structural analysis of the Severe fever with thrombocytopenia syndrome virus (SFTSV) glycoprotein-N D170N substitution and human prolow-density lipoprotein receptor-related protein 1 (LRP1).

a Proportion of 170 N variants and case fatality rate among different SFTSV genotypes from meta-review analysis. Source data are provided as a Source Data file. b Geographic distribution of 170 N variants and regional incidence of SFTSV from meta-review analysis. c Structural overview of SFTSV glycoprotein N (Gn, blue) and glycoprotein C (Gc, brown) bound to human LRP1 (pink), as derived by ZDOCK. d Zoom of the 170D in relation to proximal amino acid side chains. e Derived 170N variant in relation to proximal amino acid 172 A of LRP1.

Discussion

Our analyses revealed that highly pathogenic tick-borne viruses tend to evolve toward ecological generalism. Ecological generalism allows PTBVs to colonize diverse environments, increase transmission opportunities and enable frequent host jumps to humans. Moreover, generalist viruses are less vulnerable to extinction, as their survival is not restricted to a single ecological niche²⁴. As generalist viruses, PTBVs’ ongoing persistence and evolution are driven by a combination of intrinsic genetic adaptations and extrinsic ecological factors.

Increased selective constraints, leading to a reduction in positive selection pressure (and a lower d_N/d_S ratio) have been linked to generalist viruses that infect broad range hosts, particularly in arthropod-borne viruses^25,26. This may reflect the action of antagonistic pleiotropy, such that nonsynonymous mutations that are beneficial in one host (e.g., a vector) may be deleterious in another host (e.g., a mammal), thereby increasing selective constraints. In contrast, our study revealed that HPTBVs, which appear to be ecological generalists, exhibited an elevated d_N/d_S ratio, particularly in genes associated with cellular interactions, such as glycoprotein of Bandavirus, Orthonairovirus, Mivirus Thogotovirus Quaranjavirus, and Jingmen virus group, and the membrane, envelope, and coat proteins of Orthoflavivirus. Recent evidence that mutational fitness effects are often positively correlated across hosts suggests that multi‑host viruses can undergo localized or episodic positive selection without globally relaxed selective constraints, consistent with the elevated dN/dS ratios observed in generalist PTBVs²⁷. Such adaptive evolution was especially pronounced in three recently emerging HPTBVs—CCHFV, SFTSV, and JMTV—in which specific amino acid residues linked to the functions of the attachment to host receptors and fusion into host membrane were found to be under positive selection. Differences in the number of positively selected sites should be interpreted cautiously, as site‑specific selection inference is based on phylogenetic substitution patterns and may be partially influenced by data availability. These sites may play a crucial role in optimizing molecular adaptation to human hosts, thereby facilitating ecological generalism after spillover to humans. For example, the D170N mutation in SFTSV may enhance affinity for host receptors, potentially leading to higher incidence and increased pathogenicity. Most sequences harboring the 170D substitution were of human origin, whereas the 170 N substitution was exclusively detected in humans. These patterns are suggestive of human adaptation, particularly for 170N, which showed increased predicted binding affinity. Similar trends were observed for positively selected sites at codons 298 and 501, although the functional roles of these mutations remain to be elucidated. In addition, the predicted SFTSV glycoprotein–LRP1 complex exhibited structural stability throughout the molecular dynamics simulations, supporting the reliability of the proposed binding models (Supplementary Fig. 12). Nevertheless, further in vitro validation is required to confirm these structural interactions, particularly in human strains carrying the D170N mutation.

Multiple modeling approaches were employed to systematically evaluate ecological factors for the geographic expansion, occurrence and richness of PTBVs. As indicated from the meta-review analysis, migratory birds were susceptible to HPTBVs with geographic globalization. For instance, widely distributed TBEV, CCHFV, and SFTSV were capable of infecting bird species such as Anser, Uria, Laridae, Spatula and Hirundinidae. The dynamic dispersal patterns of these PTBVs further supported the significant role of bird migration in intercontinental dissemination, although the extent of this impact may vary depending on the virus species and their association with specific flyway indicators. However, our phylodynamic analyses did not reveal a significant association between the spatial spread of SFTSV lineages and known migratory flyways. This may be largely attributable to the limited spatial resolution of currently available global migratory bird datasets, particularly given that most SFTSV cases are concentrated in East Asia. We note that future analyses incorporating more precise, region-specific migratory bird movement data may help to better evaluate this potential transmission pathway. Of note, JMTV exhibited a positive association with flyway frequency, even though there have been no reports of JMTV infecting birds to date. Twelve tick species have been identified as potential vectors of JMTV, and seven of them are known to parasitize birds—particularly Amblyomma sparsum, Amblyomma nuttalli, and Haemaphysalis hystricis, which are among the most prevalent vector species (Supplementary Fig. 9b). It indicates that birds may not serve as direct hosts or reservoirs for JMTV but rather act as carriers facilitating the long-distance movement of infected ticks. Phylodynamic analyses could provide valuable insights into these relationships and uncover the ecological mechanisms driving PTBV dissemination.

Animal hosts are not only critical for the local persistence of PTBVs but also serve as key drivers for the emergence of novel viruses. This is evidenced by their significant contribution to PTBV occurrence (as shown by BRT models) and richness (as shown by GAM models) in humans. Regions with high animal diversity and density, particularly livestock, may act as incubation zones for PTBVs. These findings align with previous research on the single tick model, Haemaphysalis longicornis, where viromes influenced by surrounding animal diversity led to increased PTBV emergence in nearby human populations. The importance of animals is especially pronounced in hotspot regions, where a strong linear correlation was observed between PTBV richness in animal hosts and nearby human populations. This highlights animal hosts as incubators for the frequent emergence of novel PTBVs. However, the lack of PTBV records in animal hosts, whether from meta-reviews or sequencing data, limits further evidence gathering. This highlights the need for enhanced surveillance of PTBV genetic diversity in animal hosts, especially in hotspot regions like northeastern Asia, to better understand the ecological mechanisms driving PTBV emergence.

This study had several limitations. First, the ecological variables, such as climate factors used in our models, did not fully align with the timing of reported PTBV occurrences. Using average ecological predictors may have obscured temporal variations. Pooled prevalence estimates derived from random‑effects models are intended as descriptive, group‑level summaries of detection and exposure patterns rather than precise measures of infection risk. It should be interpreted cautiously as hypothesis‑generating, given the heterogeneity and limited control of confounding across source studies. Second, the selection of pseudo-absence locations could include false negatives. Sensitivity analyses showed that model performance was only moderately affected by pseudo-absence sampling, with low false positive rates (0.0157 [0.0131, 0.0184]) and training–testing AUC differences below 0.05 on average (0.00459), indicating limited overfitting and minimal bias introduced by pseudo-absence uncertainty. Lastly, the limited genomic data for some newly reported LPTBVs and NPTBVs may have biased our evolutionary analysis, as their shallow phylogenetic trees reflect lower viral diversity.

Our analysis documented the ecological traits of TBVs associated with their shift to generalist viruses. The Ecological Generalism Index (EGI) proposed in this study provides a standardized, integrative measure of ecological generalism by incorporating host, vector, and environmental dimensions, enabling comparative assessment of ecological breadth and associated risk across TBVs. The significant association between EGI and pathogenicity, together with its robust predictive performance, suggests that EGI may be useful for identifying TBVs with elevated spillover or emergence potential. To enhance its practical utility, compiling viruses with high EGI but currently low documented pathogenicity (Supplementary Table 1) may help prioritize candidates for surveillance and illustrate how the EGI framework can inform early warning and One Health–oriented monitoring strategies. Despite these strengths, several limitations should be acknowledged. Surveillance efforts are often biased toward human cases, while animal hosts and tick vectors—particularly in under-sampled regions—remain comparatively under investigated. Such sampling imbalances may influence EGI construction and downstream modeling, potentially affecting estimates of ecological breadth. In addition, detections of TBEV, CCHFV, and SFTSV were analyzed at the lineage level, whereas other TBVs were treated as single viral entities due to limited strain- or lineage-level resolution. This aggregation may lead to overestimation of ecological generalism if genetically distinct lineages with narrower host ranges are combined. Collectively, this study highlighted the utility of extensive data sets that integrate epidemiological and genomics data to reveal the ecological and evolutionary factors associated with the increased spread of PTBVs. Expanding similar research to other generalist viruses could provide broader insights into the evolution of virus spillovers, particularly for those viruses with the greatest impact on public health.

Methods

Ethics statement

This study was based solely on publicly available sequence data retrieved from open-access databases. No human participants, animals, or identifiable personal data were involved; therefore, ethics approval was not required according to institutional and national guidelines.

Occurrence and pseudoabsence data

We assembled a global occurrence database for 32 human pathogenic tick-borne viruses, utilizing online records from various sources including PubMed, Web of Science, GenBank, and the Sequence Read Archive (SRA) (Supplementary Fig. 1): Kyasanur forest disease virus/KFDV, Powassan virus/POWV, Tick-borne encephalitis virus/TBEV, Omsk hemorrhagic fever virus/OHFV, Crimean-Congo hemorrhagic fever virus/CCHFV, Severe fever with thrombocytopenia syndrome virus/SFTSV, Heartland virus/HRTV, Colorado tick fever virus/CTFV, Jingmen Tick Virus/JMTV, Haseki tick virus/HSTV, Alongshan virus/ALSV, Beiji nairovirus/BJNV, Tamdy virus/TAMV, Songling virus/SGLV, Dugbe virus/DUGV, Wetland virus/WELV, Erve virus/ERVEV, Yezo virus/YEZV, Avalon virus/AVAV, Tacheng Tick Virus 1/TCTV-1, Nairobi sheep disease virus/NSDV, Bourbon virus/BRBV, Oz virus/OZV, Thogoto virus/THOV, Dhori virus/DHOV, Quaranfil virus/QRFV, Guertu virus/ NSDV, Bhanja virus/BHAV, Tacheng Tick Virus 2/TCTV-2, Eyach virus/EYAV Nuomin virus/NOMV), and Xue Cheng virus (XCV). This process involved two independent reviewers (W.G.P. and Z.N.Q.) who extracted peer-reviewed publications by searching PubMed and Web of Science for studies published before Jan 31, 2025, using the names of each viral species as search terms.

The inclusion criteria were as follows: laboratory-confirmed detection of PTBVs; availability of ticks or host source information; availability of sampling geographical information. More detailed inclusion and exclusion criteria can be found in Supplementary Fig. 1. The database includes various types of infection events related to PTBVs, such as scientific names or common names of ticks and animal hosts; collection locations and sampling date; positive rates or numbers for PTBVs in ticks, animals, and humans, while tick surveillance data were standardized using the minimum infection rate (MIR) to avoiding overestimation of infection frequency²⁸; detection methods, including molecular assays, pathogen isolation evidence, or serological assays in animal hosts. Comparisons of pooled prevalence between groups were performed using random-effects models to account for between-study heterogeneity, using the R package meta v8.2–1.

All viral sequences and related meta-information were obtained from GenBank (https://www.ncbi.nlm.nih.gov/genbank/) by searching for viral species names and alternative names up to June 01, 2024. Meta-transcriptomic reads for ticks were downloaded from the SRA, excluding those derived from laboratory settings or focused solely on tick organs. PTBV-positive records from the SRA were identified through downstream analysis. After filtering records lacking detailed geographic information, we integrated the metadata from the databases described above to construct a comprehensive occurrence database for PTBVs. Pseudo-absences were randomly sampled within a 30–3000 km radius based on known tick or host dispersal ranges around the occurrence locations to achieve a 1:3 ratio^29–31. We conducted sensitivity analyses by repeating random pseudo‑absence generation 100 times to evaluate the robustness of model performance to pseudo‑absence uncertainty.

We calculated the Ecological Generalism Index (EGI) for each PTBVs based on their prevalence in ticks and animal hosts, tick and animal diversity, and ecological environments Supplementary Fig. 9a). To further address sampling imbalance, EGI was re-estimated at the lineage level for three overrepresented viruses, including CCHFV, JMTV, and SFTSV, improving comparability across taxa. Given empirical evidence that the natural transmission cycle of tick-borne viruses critically depends on the simultaneous presence of competent vectors, susceptible hosts, and suitable ecological environments, none of these components can be considered dispensable. Therefore, equal weighting was adopted to reflect their complementary and interdependent roles in sustaining viral circulation.

The EGI cut-off to classify viruses as HPTBVs (positive class) and LPTBVs (negative class) was determined based on a traversal of the EGI value of each virus. The threshold achieving the highest balance accuracy, i.e., the average of true positive and true negative rates, is the cut-off. A sensitivity analysis was performed to assess the robustness of the EGI, and added ROC-based performance evaluation to quantify the EGI discriminatory ability (Supplementary Fig. 9b).

Eco-climate predictors

Climate data were obtained from version 2.1 of the WorldClim database (https://www.worldclim.org/), which includes six climatic variables: average temperature, minimum temperature, maximum temperature, precipitation, wind speed, and solar radiation, collected from 1970 to 2000. We calculated the mean of these climatic variables over their respective time spans to use as predictors for ecological modeling. This approach may introduce a temporal mismatch with the exact dates of viral emergence, which remain uncertain. Biodiversity data for mammal hosts and domestic animal density were sourced from the Global Mammal Assessment database (https://globalmammal.org/) and the Gridded Livestock of the World database (https://dataverse.harvard.edu/dataverse/glw/), respectively. Ecotypes were obtained from the EarthEnv (https://www.earthenv.org/). We summarized the headcount of domestic animals (poultry, goats, buffalo, cattle, sheep, and pigs) into a single predictor, using the Shannon index to represent livestock mammal biodiversity, conducted by the Python package “skbio” (http://scikit-bio.org/). We calculated the Shannon diversity index for each area at a 10 km × 10 km resolution based on the prevalence of 12 land-cover types, using the Python package scikit-bio (“skbio”). This metric was defined as the variable ecotype diversity. The average annual international flight frequency between countries (1970–2021, World Data Bank, https://data.worldbank.org/) was used as an indicator of human mobility and travel intensity. We included the import and export quantities of major livestock species—including goats, swine/pigs, chickens, ducks, geese, turkeys, horses, asses, mules and hinnies, camels, rabbits and hares, cattle, buffalo, and sheep—obtained from FAO statistics (1961–2023, https://www.fao.org/) to represent global livestock trade and movement. Additionally, bird migratory flyways were downloaded from the East Asian-Australasian Flyway Partnership (EAAFP, https://eaaflyway.net/the-flyway/).

Meta-transcriptomic sequence data, including self-obtained and downloaded data sets, were analyzed according to a previous approach to obtain the virome diversity (Shannon index) and PTBVs relative abundance (Reads Per Million) as predictors for ecological modeling²⁰. The following families of RNA viruses were analyzed: the Phenuiviridae, Rhabdoviridae, Flaviviridae, Chuviridae, Sedoreoviridae, Nairoviridae, Orthomyxoviridae, and Peribunyaviridae. We assigned all the above predictors to the occurrence and pseudo-absence data based on their coordinates, using the closest distance to the data grid from the downloaded predictors database.

Ecological modeling for PTBVs occurrence

To explore the relationship between the probability of PTBV presence within each 10 × 10 km geographic region and the environmental, ecoclimatic, and biological predictors known or hypothesized to influence the ecological suitability of PTBVs and HPTBVs, we compared the model performance of three ecological modeling approaches: Boosted Regression Trees (BRT), Random Forest (RF), and Support Vector Machine (SVM). BRT, RF, and SVM were implemented using the ‘sklearn’ package in Python. Aside from the internal cross-validation (75%–25% split) of the modeling procedure, both presence and pseudo-absence points were divided into an 90% training set and a 10% testing set for each sub-model, with model performance evaluated on the withheld test dataset. The best model was then selected to assess the importance of all included predictors for PTBV occurrences.

Generalized additive model (GAM) for PTBVs richness

GAM was employed to investigate the effects of ecological variables, such as average temperature, precipitation, solar radiation, ecotypes, mammal biodiversity, livestock biodiversity, bird flyway number, human population, tick virome diversity, PTBVs abundance, the number of PTBV species in tick or animal hosts, on the number of PTBV species of PTBVs in humans within the study grid (10 × 10 km). As most recent emerging PTBVs occurred in northeastern Asia, the data set located in Asian regions was extracted and randomly divided into a training set comprising 90% of the data and a leave-out testing set constituting the remaining 10%. A variance inflation factor (VIF) greater than 10 was applied to eliminate factors exhibiting collinearity. The overall intersectional and specific individual effects of the selected predictors were modeled using GAM to assess the number of PTBV species in humans within the study grid, facilitated by the Python ‘sklearn’ package. Five-fold cross-validation was utilized for model optimization on the training set.

Analysis of selection pressures on TBVs

For each virus analyzed, we removed duplicate sequences with a threshold of 100% nucleotide similarity using CD-HIT v4.8.1³², selecting the sequence with the earliest sampling date and a length greater than 50% of the complete gene length as the representative. Representative amino acid sequences from the same virus species were aligned for each gene by GLProbs³³. Codon-based sequence alignments were performed on coding region sequences, guided by their corresponding amino acid alignments using Pal2Nal v14.1³⁴. We checked recombination events using RDP4³⁵ and removed the recombinants from the downstream analysis. A genus-level nucleotide alignment was constructed by stepwise merging the species-level alignments using MAFFT v7.490³⁶, with ambiguously aligned regions removed using TrimAl v1.2³⁷. Phylogenetic trees were inferred from the alignments using a maximum likelihood (ML) approach implemented in IQ-Tree v2.3.6³⁸. The best-fit nucleotide or amino acid substitution model was selected for each multiple sequence alignment, and the ML phylogenetic tree was assessed with bootstrap tests (1000 replicates). All species-level phylogenies were rooted using outgroup sequences as previously reported.

To estimate selection pressures, we estimated the number of nonsynonymous (d_N) and synonymous (d_S) substitutions per site along the tree branches using a codon substitution model³⁹, with individual parameters based on the MG94xHKY85 substitution model implemented in HyPhy v2.5.63⁴⁰. To help account for the presence of transient deleterious mutations, in species-level trees, we determined the d_N/d_S ratio for internal and leaf branches separately within each TBV species^41,42. In addition, we computed d_N/d_S for individual codons across all sequence alignments using the FEL⁴³, FUBAR⁴⁴ and MEME⁴⁵ methods. Finally, d_N/d_S ratios were compared between groups using the Mann-Whitney U-test.

Phylogeographic modeling on PTBVs dispersal

To evaluate the presence of a temporal signal in the phylogenetic data, a linear regression of root-to-tip distances against sampling dates on maximum likelihood (ML) phylogenetic trees was conducted by using TempEst v1.5.3⁴⁶. A moderate to strong temporal signal was indicated by a positive correlation (R² = 0.52 for TBEV Siberian subtype (TBEV-Sib), 0.44 for TBEV Far Eastern subtype (TBEV-Fe), 0.33 for TBEV European subtype (TBEV-Eu); 0.83 for CCHFV Asia-Africa lineage, 0.53 for CCHFV Europe-Africa lineage and 0.56 for SFTSV). As JMTV didn’t show a significant temporal signal, such that sequences do not evolve according to a molecular clock, the fixed topology was used for the downstream analysis. The dispersal history of PTBV lineages was estimated using discrete space phylogeographic methods implemented in BEAST v 1.10.4⁴⁷ with the BEAGLE library v4⁴⁸. Phylogeographic reconstruction was conducted under a generalized linear model (GLM) to assess the impact of hypothesized predictors on the dispersal of PTBV lineages⁴⁹. This included an HKY nucleotide substitution model with gamma-distributed among-site rate heterogeneity, a Bayesian Skygrid coalescent prior on population size, and an uncorrelated relaxed clock. Multiple potential predictors, including average temperature, precipitation, animal diversity, livestock diversity, human population, flyway numbers, and ecotype diversity, were incorporated into the GLM-diffusion phylogeographic model. We ran five million steps for at least one chain, sampling every 5000 steps. The first 10 to 20% of the chains were discarded as burn-in. MCMC convergence was assessed using Tracer v.1.7.1⁵⁰, ensuring that effective sample sizes for continuous parameters were greater than 100.

Protein structure modeling

The three-dimensional structures of the SFTSV glycoprotein and human LRP1 were obtained from the Protein Data Bank (PDB; accession number: 8ILQ) and UniProt (accession number: H0YJ88), respectively. The D170N amino acid substitution in the SFTSV glycoprotein structure was introduced using iCn3D⁵¹. The interaction between human LRP1 and the SFTSV glycoprotein was predicted through docking analysis performed with ZDOCK v3.0.2⁵². Comparative homology modeling was also carried out using iCn3D⁵¹.

Reporting summary

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

Author contributions

L.-T.T. conceived, designed, and supervised the study. C.-W.C. and J.-N. designed and supervised the meta-virome study from self-collected samples. N.-X.B., W.-G.P., W.-Y.Q., Z.-J.J., and Z.-N.Q. searched, screened, and assessed the publications. L.-X., X.T., X.-T., A.-T.F., C.-Z.Y., and C.-D.S. helped with checking the meta-review database. N.-X.B., W.-Y.Q., and Y.-Y.T. searched, screened, and assessed the NCBI and SRA databases. N.-X.B., Y.-Y.T., S.-Y., and C.-Y.X. extracted and analyzed the data and verified the underlying data of the study. N.-X.B., Y.-Y.T., and W.-G.P. helped with checking data and constructing the figures. N.-X.B. wrote the drafts of the manuscript and interpreted the findings. L.-T.T., C.-W.C., J.N., E.C.H., F.-Z.Z., and Y.-Y.T. commented on and revised drafts of the manuscript. The corresponding author had full access to all the data in the study, accepted responsibility for the decision to submit for publication, and approved the final version.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The meta-transcriptomic data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) database under BioProject accession code PRJNA841744. The meta-review underlying this study was registered in PROSPERO under registration number CRD420251018201. The data included in the meta-analysis align with the study selection, extracted variables, and analytical datasets specified in the PROSPERO registration and the registered protocol. The raw data are subject to restricted access due to data privacy regulations. Access to the data can be obtained upon request to Prof. T.T. Lam (email: ttylam@hku.hk). Data will be made available for scientific research purposes only, subject to approval and, if required, the signing of a data use agreement. Requests for access will typically be reviewed within two weeks. Once access is granted, the data will be available for a period of one month. The processed data are provided as source data in this paper. The structural data used in this study are available in the Protein Data Bank under accession code 8ILQ. The receptor protein sequence is available in UniProt under accession code H0YJ88. Source data are provided with this paper.

Competing interests

The authors declare no competing interests

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-72514-5.