This study revealed new insights about the evolutionary dynamics of the FMDV’s global transmission dynamics at serotype level. The most likely country of origin for each serotype was identified, along with its historical spread characteristics, and divergence patterns across its historical dispersal. Finally, we assessed the impact of each host interaction in the spread of FMDV, providing a comprehensive characterization of transmission dynamics between host species.
Phylogeographic patterns of FMDV spread
Global patterns of FMDV spread were considerably asymmetric in its spatiotemporal arrangement, showing important variation among all serotypes, as previously observed by  and . On the other hand, our results yielded discrepancies regarding the phylogenetic relationships of FMDV serotypes due to the disagreements observed in the cladistic characterization of FMDV serotypes (monophyletic or polyphyletic origin) .  and  suggested that O, A, Asia1, C, and SAT3 were monophyletic, while SAT1 and SAT2 serotypes were polyphyletic. However, our results indicated the presence of only three monophyletic serotypes (O, A, and Asia1), whilst all SAT serotypes appeared to have multiple ancestral origins which can be related to multiple points of independent introduction of the virus.
Serotypes with global distribution (O and A)
Serotype O has shown a remarkable widespread distribution across the globe. In half of a century, this serotype reached almost all continents, causing dramatic economic losses [58, 80, 81]. Root state posterior probability analysis inferred Belgium as the most likely center of origin for this serotype, which, as a result of being responsible for the majority of outbreaks worldwide , we can observe multiple centers of diversification in most of the continents. Our phylogeographic analysis showed that this spread has been characterized by lineage dispersal events between distant regions (i.e. to regions not sharing international dry borders with the origin country), instead of dispersal events between neighboring countries, which may be one of the keys for its successful global spread. Bayesian skyline plot showed a severe decline in the genetic diversity around early 21th century, this interesting pattern also observed by . This decline and recover in the effective population size could be directly related to the increase in the FMDV outbreaks that occurred worldwide, which was followed by an intensive control and prevention strategies. The intense wave of outbreaks occurred during that period worldwide included countries such as, Argentina [83, 84], the United Kingdom [85, 86], Brazil , India , and Taiwan [89, 90]. As we observed in our phylogeographic visualization, there is strong evidence that most of these outbreaks were strongly interconnected [25, 91–93], evidencing local and long-distance spread of serotype.
One of the reasons for the success of the evolutionary diversification of this serotype may also be related to the diversity of hosts that it affects, which is the highest among all serotypes. Globally, B. taurus represented the most important host species for the spread of serotype O, while S. scrofa was mostly related to the spread of this serotype in southeastern Asia. Thus, phylodynamic analysis suggested that viral transition rate between these two livestock was the strongest reported between all the reported hosts.
Following the pandemic patterns showed by serotype O, the next large-scale potential of diffusion was exhibited by serotype A. Phylogeographic analysis suggested India as the most likely center of origin of the current circulating serotype A strains. Supporting previous studies [26, 60], we observed that India was also a key source of dispersal events for this serotype since most of the current strains are strongly related to India. Whole genome sequences of this serotype have been recorded in three continents, Asia, Africa, and South America, where it was reported as the causing agent of one of the biggest FMDV outbreaks, which occurred in Argentina in 2011, affecting a total of 2,126 herds . It is important to note, that there is evidence of a posterior spread of these serotypes (O and A) to other countries, mainly in South America, since both of them are currently found in nearly every country of the continent [26, 82]. However, due to the lack of whole-genome data, we were unable to further assess this spread.
As expected, the main host affected by serotype A was B. taurus. This species had an important role in its viral spread , especially in this globalized era, where the continuous increase in livestock trading markets facilitates the spread of transboundary animal diseases . Likewise, our phylodynamic analysis showed that the most intense host species transmission route occurred from B. taurus to S. caffer, and apparently, the reverse transmission is an infrequent event.
Asia1 and SAT serotypes
Whereas our results showed that serotypes O and A have spread worldwide, serotypes Asia1 and SATs remained non-pandemic and confined in their endemic regions [82, 95]. Since there is a lack of detailed sequences data available, especially for African countries, it is important to note that these results may vary with a better representation of the currently circulating virus, although they support what has been previously described [59, 78, 79, 82].
Undoubtedly, India has been historically considered as one of the most important countries for the spread and maintenance of FMDV, especially for serotypes A, and Asia1 [26, 56, 79, 82] Indeed, our phylogeographic analyses showed India as the most likely origin country for Asia1 serotype [26, 79]. The spread of this serotype was mainly restricted to Asia [53, 56, 78], and characterized by local movements across the neighboring countries surrounding India, China and Malaysia, where it is well known that free and unrestricted animal movements across country borders may play a key role in the spread of FMDV [56, 96]. We also observed India as a key center of dispersal for this serotype, which coincides with previously reported results . The arrival of Asia1 into Turkey in 2013 represents one of the most recent and longer dispersal events reported for this serotype, which was directly related to an Indian sub-lineage of the virus . Likewise, there have been sporadic incursions into other countries such as Greece in 1984 and 2000 , Malaysia in 1999  or Turkey in 2017 , whose outbreaks seemed to be caused also by independent sub-lineages from the rest of the outbreaks observed in these regions.
Despite a previous study described multiple potential origins for SAT serotypes, (i.e., SAT1 in Zimbabwe and SAT2 in Kenya ), our root state posterior probability results suggested Uganda as the most likely origin for all of them. Likewise, our phylogeographic analysis also highlighted the importance of Uganda as a primary source of dispersal events to different countries, where the most strongly significant routes were found from Uganda to Nigeria (SAT1), from Uganda to Gaza strip (SAT2) and from Zimbabwe to Botswana (SAT3).
SAT serotypes (SAT1, SAT2, and SAT3) are characterized by a higher proportion of local spread, limited across their endemic areas. This spread occurred mainly in southeastern Africa, where nomadic pastoralism across international borders and animal trade in the sub-Saharan region is one of the most practiced forms of livestock movements [48, 55, 78, 79]. These results complement the observations made by , who highlighted how nomadism and transhumance play a key role in disease transmission, especially in African countries.
Previous studies have highlighted the importance of African buffalo (Syncerus caffer), hypothesizing that current FMDV genotypes may emerge in domesticated host species from viral reservoirs maintained by this species [49, 53, 59, 79, 99–105] However, the uncertainty over the involvement of African buffalo arose the need for deeper research to confirm its influence in livestock outbreaks  Our results coincide with the evidence mentioned in a recent study by , where cattle appeared as the most important host species for the spread of FMDV, while buffalo played a secondary role. This pattern was observed not only in SATs but in all serotypes studied.
In general, we observed considerable differences in the spatiotemporal dynamics exhibited by the different serotypes. Where the serotypes with global distribution (O and A) presented the most asymmetrical pattern in the annual genetic diversity in comparison with (SAT and Asia1 serotypes). Cattle was observed to play a key role in the historical spread of all serotypes of FMDV. Likewise, our phylodynamic analysis inferred that the transmission route from cattle to buffalo was the most highly supported, pattern that was also observed for all serotypes, independently of its spread potential.
Serotypes such as Asia1 and SATs presented local spread rates, mainly associated with cattle and sheep (with special importance of buffalo in the case of SATs serotypes) supporting previously described results [48, 79], while serotypes O and A showed long-distance spread, covering higher extensions of territory between each outbreak, which also confirms previously described information . These serotypes presented the highest variety of susceptible hosts, although we speculate that the main reason for their successful long-distance spread relies mostly on the international movement of cattle and swine due to the intensive commerce between countries.
Finally, important limitations related to the use of whole genome relay in the lack of good global data, especially in African countries which remains endemically affected by five different serotypes, therefore some countries with known FMDV circulation are not part of this study. However, to reduce the bias generated by the strong unbalance of the available data in both dimensions (i.e., number of samples per country and uneven number of samples per host species), we removed all the sequences that where duplicated (i.e., represented the same outbreak multiple times), which, in the case of big outbreaks such as United Kingdom 2001/2007, Argentina 2001, and Japan 2010, accounted for hundreds of sequences representing each event. This limitation is common among phylogenic studies with no yet best alternative, this is true mainly because sample that are available hosted in public databases or from diagnostic laboratories . Although whole genome sequences are increasingly proving to be a more accurate tool for phylogenetic analyses [107, 108], its high cost in comparison to studies considering partial genome results in lower availability of WGS, which became the major limitation for the construction of our dataset, resulting several countries with known reports of FMDV but lacking genetic data. Finally, it is important to highlight that, despite the nucleotide sequences encoding the capsid protein VP1, VP2, and VP3 are sufficient to identify FMDV at serotype level, WGS is preferred because of its higher accuracy in the determination of the genetic relationship between the reported cases .
Studies considering whole genome sequences should be preferred over partial sequence research to ensure the importance of considering virus spread in its overall context
[53, 107, 108]. Besides, the growing awareness of the importance of using whole genome sequences to assess the evolution of infectious diseases, and more specifically for RNA viruses as FMDV plays a key role on the future ability to analyze the ever-increasing volume of data accurately, getting closer to a real-time assessing of disease outbreaks . However, the use of whole genome sequences represented a limitation in our study since the lack of FMDV sequences in a given country does not mean that the virus has not been circulating in that country but maybe associated with technical or economic constraints, therefore interpretation requires caution due to the possible introduction of sampling bias. The popularization of whole genome sequencing will help not only to increase the available information about the virus, but also have a direct impact on promoting new and more specific measures for disease control [24, 59, 110]. The result of such improvement in disease surveillance would not only be beneficial for the targeted region, but also for all the areas that are directly connected (i.e., through geographical limits) and indirectly (i.e., through commercial networks), including countries currently considered as free zones . Even though past studies have found recombination in WGS of FMDV [53, 60], it is important to mention that after filtering the dataset to keep just one sequence per outbreak event, we did not found evidence of recombination on the remaining samples.