Migratory Whooper Swans Cygnus cygnus Transmit H5N1 Virus between China and Mongolia: Combination Evidence from Satellite Tracking and Phylogenetics Analysis

In late 2014, a highly pathogenic avian influenza (hereafter HPAI) H5N1 outbreak infected whooper swans Cygnus cygnus wintering at the Sanmenxia Reservoir area, China, and raised concerns about migratory linkages between wintering and breeding grounds of whooper swans. In this study, 61 swans were satellite tracked from 2013 to 2016 to determine the spatial association of their migration routes and H5N1 outbreaks, and 3596 fecal samples were collected along the migration routes for virology testing. Swans departed the wintering grounds and migrated along the Yellow River, and flew over the Yin Mountains in China. The Brownian bridge movement model showed there was a high degree of spatiotemporal overlap between the core use area along the spring migration pathway and historical H5N1 events in China and Mongolia from 2005 to 2015. The H5N1 strain was isolated and phylogenetic analyses confirmed that the HA gene sequence generated is genetically similar to that of the epidemic strain at a previous wintering site (the Sanmenxia Reservoir area) along its flyway. Our results identified a previously unknown migratory link of whooper swans in central China with Mongolia and confirmed that the swans could carry the HPAI H5N1 virus during migration, resulting in long-distance transmission.

Departure dates from the wintering grounds ranged from 17 February to 27 March (median 7-8 March, n = 63). Arrival dates at the breeding sites in Mongolia ranged from 27 February to 23 May (median 9-10 April, n = 22).
Whooper swans migrated along the Yellow River of Shanxi, Shaanxi and Inner Mongolia and its tributaries (including the Yan River, Hekou Reservoir, Tuwei River), flew over the Yin Mountains in China, and then flew to the rivers and lakes of central-northern and western Mongolia. We summarized the spring migration routes into two stages as follows: From the wintering grounds to the Yellow River stretch in Inner Mongolia in China. The majority of the whooper swans (n = 57) departed from the Sanmenxia Reservoir area and flew northwest directly to the Yumen Estuary. With a short stopover (mean 3.0 d, n = 31) period, they migrated north along the Yellow River, then they took a longer stop at the Yulin wetlands (mean 4.3 d, n = 15) and Ordos wetlands (mean 8.5 d, n = 12) before arriving at the Yellow River stretch in Inner Mongolia for their longest stay (mean 24.0d, n = 43) Some whooper swans took tributary migration routes. Eight swans departed from the Sanmenxia Reservoir area and migrated west along the upstream of the Yellow River; two swans traveled directly north by way of Linfen and Lishi; fifteen swans departed from the Yumen Estuary (n = 6) and the Yulin(n = 4) and Ordos (n = 5) wetlands following a northwest route that bypassed the Maowusu Sandy Land and Kubuqi Desert, and then they finally arrived at the Yellow River stretch in Inner Mongolia.
Departing from the Yellow River stretch in Inner Mongolia to the breeding sites in Mongolia. The majority of the whooper swans (n = 31) departed north directly from the Yellow River stretch in Inner Mongolia in China to Mongolia. One whooper swan migrated west by way of Shizuishan, Alxa Left Banner, Jincang and Zhangye. Migration trends in Mongolia mainly covered three routes.
Route 1: Approximately half of the whooper swans (n = 16) flew directly north to the river basins in central-northern Mongolia. Of these swans, two remained at Ongiyn Gol for the entire summer, eight stayed near the Orkhon River, Hanuy Gol and Selenga River, and three moved to the northernmost lakes of the Shinshkhed River basin (including Khovsgol Lake and Dood Tsagaan Lake). Only one swan migrated west along the northern Khangai Mountains into Khyargas Lake in western Mongolia after a short stay at the Orkhon River basin.
Route 2: Eleven swans used a northwestern route along the valley between the Khangai Mountains and Govi-Altayn Nuruu. Of these 11 swans, four spent the summer in the Baydrag Gol (including Buuntsagaan Lake) and the Tuin River (including Ulungur Lake). Two stayed at Buuntsagaan Lake and Ulungur Lake for short periods (5 d and 7 d, respectively), and then, they tracked northeast across the Khangai Mountains and into the Orkhon River basin. Five swans continued to migrate along the Zavkhan River before they arrived at the Khovd River basin (Harus Nur, Uvsiin Khar Us Lake, Khara Us Nur and Khara nor) in western Mongolia, even migrating to the boundary between Russia and Mongolia. (Fig. 1).
Route 3: A small proportion of swans (n = 3) traveled northeast and retraced their paths to Tuul Gol for the summer.
Key stopovers and breeding sites. As shown in Table 1, most whooper swans stayed at the Yellow River stretch in Inner Mongolia from 20 February to 6 May (n = 61), the Yumen Estuary from 27 February to 28 March (n = 36), the Ordos Lakes from 25 February to 2 May (n = 23), the Hekou Reservoir from 25 February to 7 April (n = 16), and Hongjian Nur from 19 February to 15 April (n = 14). More than 10 whooper swans stopped at these sites and stayed for 3 days or longer. Thus, we identified these sites as the key stopover sites during the spring migration of whooper swans in China.
In Mongolia, breeding or molting sites of whooper swans ranged from the westernmost Khovd River basin to the central Gobi wetlands of Ongiyn Gol, Baydrag Gol and the Tuin River, as well as northern-central basins of the Orkhon River, Tuul River and Selenga River, and the northernmost Shinshkhed River basin. Throughout these river basins, some lakes and rivers served as important sites with higher location densities or longer durations for swans during breeding or molting periods based on the tracking data (Table 1). Spatiotemporal analysis of BBMM and H5N1 outbreak. BBMM shows that whooper swans wintering at the Sanmenxia Reservoir area occupied more space in China and Mongolia (Fig. 2) during the migration. At the sites that experienced H5N1 event outbreaks, in May 2015, several hundreds of wild fowl including wild swans and ducks that died were infected with H5N1 at Wuliangsu Lake and Tsagaan Nur, areas that sit exactly on the core use area of the 95% contour level for whooper swans and where there was a high population density distribution zone that was exactly identical with the migration data of whooper swans. Similarly, we also found there was a high degree of spatiotemporal overlap between the core use area of the 95% contour level for the spring migration pathway and historical H5N1 events including Doitiin tsagaan Lake, Khunt Lake, Saikhan Soum, and Erhel Lake in Mongolia, which sit on central-north river and lake wetlands, from 2005 to 2010 and exactly in the breeding period of whooper swans on the basis of tracking date.
We not only analyzed the relationship between the H5N1 events and home range but also evaluated the risk factors in epidemic diffusion. There was a significant difference in the distribution of the H5N1 events and random points in 95%, 99% and 100% of the home ranges (P < 0.01). It can be seen from Table 2 that there is a substantial relationship between the home range and the H5N1 events (OR > 1). The probability of outbreaks in 95%, 99% and 100% of the home range is respectively 57.00, 2.33 and 1.67 times greater than the probability of an outbreak outside the 95%, 99% and 100% of the home range. analyses indicated that the HA segment of the strain belongs to clade 2.3.2.1c and has a high genetic homology with that of viruses previously isolated from whooper swans wintering in the Sanmenxia Reservoir area 21 and a virus isolated from a black-necked grebe at Inner Mongolia (Fig. 3). This result suggests that whooper swans may transmit H5N1 viruses from the wintering site to stopover sites and possibly to breeding sites.

Discussion
The population density of whooper swans at the Sanmenxia Swan Lake in the National Urban Wetland Park increased rapidly between 2010 (1.58 individuals/ha) and the winter of 2014 (32.11 individuals/ha). However, the subsequent decline to 13.92 individuals/ha during the wintering survey in December 2015 suggested that the population density may have been influenced by the AIV outbreak during the winter of 2014. The whooper swan was the dominant species at Swan Lake because of artificial provisioning with corn. The increased density of the whooper swans at the artificial feeding area greatly increased the chance of AIV spreading rapidly among individuals 2,3,22 . Therefore, it is believed that the relationship between population density and the size of feeding areas is important for H5N1 outbreaks. In addition, the total number of swans at the Sanmenxia Swan Lake in the National Urban Wetland Park in December 2015 was 24.3% lower than in 2014. This dynamic was similar to the population dynamics of the bar-headed geese Anser indicus that were infected by H5N1 in the 2005 breeding season, and the population declined significantly at Qinghai Lake in the 2006 breeding season 23 . The intraspecific encounter probability increased with higher population densities, which carries a higher risk of spreading AIV, resulting in the outbreak of avian influenza, and subsequently, the encounter probability declined with a lower population density due to the outbreak of avian influenza, which may have been the result of population self-regulation. The swans had to balance their population dynamics through a "population increase-disease outbreak-population decrease" mechanism.
Our detailed summarization of the spring migration routes of whooper swans wintering in the middle reaches of the Yellow River in China and their summer breeding areas in central-northern and western Mongolia provides new baseline data for whooper swans in eastern Asia. Interestingly, we found migratory swans tended to use a narrower flyway during their earlier period of spring migration and subsequently dispersed when they arrived on breeding sites 16 . After departing the Sanmenxia Reservoir area, the majority of satellite-marked swans migrated intensively along the mainstem of the Yellow River and occupied more key stopover sites for up to 3 months. The minority of swans that flew across landscape barriers, such as the Kubuqi Desert, Maowusu Desert and Luliang Mountains, had fewer stopovers and a shorter migration time.
However, after a long period along the Yellow River in Inner Mongolia, swans expanded their migration course by flying to the central Orkhon River, Selenga River basin, and the northern Shinshkhed River basin as well as the western lakes along the Zavkhan River between the western Gobi Altai Mountains and Hangay Mountains, which cover a large area of Mongolia. We suspect that this funnel-like migratory trend may increase the risk of H5N1 dispersal.
Of the 61 tracked swans, 58 selected the upper reaches of the Yellow River in Inner Mongolia for their stopovers, where they foraged and rested for the longest period (24.74 ± 9.38 d). In March 2015 and April 2016, we observed that there was a large area of intensive farmland along the Yellow River that tens of thousands of swans occupied. In March and April every year, this farmland that consists of crops including sunflower and corn is full of water from the Yellow River and forms a flood plain that is favorable for whooper swans to rest and forage. Additionally, the wetlands are covered with reeds that provide good shelter for wild waterbirds and experience less human disturbance, which is likely why swans preferred to remain here for extended periods.
Takekawa et al. analyzed the spring migration routes of waterfowl wintering at Poyang Lake, China, using satellite tracking data 24 . Of the eight tagged birds, one falcated teal Anas falcata flew to the upper reaches of the Yellow River in Inner Mongolia instead of following the main migration routes crossing the Bohai Sea to Russia, which is one of the important stopovers of whooper swans. Therefore, it appears that some waterfowl wintering at the Yangtze River migrate to the upper reaches of the Yellow River and aggregate with whooper swans. The waterfowl using the Central Asia Flyway, such as bar-headed goose from Qinghai Lake, flew to breeding sites in central Mongolia by also crossing whooper swan migratory stopover locations such as Juyan Lake in western Inner Mongolia 25 . A total of 30,000 waterbirds, including whooper swans, whistling swans Cygnus columbianus, bar-headed geese, ruddy shelducks Tadorna ferruginea, and swan geese Anser cygnoides, were recorded during the survey of the upper reaches of the Yellow River from April 2015 to March 2016. This shows that the Yellow River in Inner Mongolia is the key convergence zone where waterbirds from eastern China and central Asia aggregate, indicating a high risk of H5N1 spread and evolution 26 .
It should be noted that the HA gene segment of the virus isolated from domestic ducks in Zhejiang and Jiangsu in eastern China was even more closely related to the virus isolated in this study from wild waterfowl in the upper reaches of the Yellow River. This indicates possible cross-species transmission. The whooper swans are widely distributed in eastern China, such as in the Zhejiang and Jiangsu Province 27 , where there exists considerable backyard and free-grazing poultry production. The infected waterfowl and whooper swans could have migrated to the Yulin and Ordos regions, the upper reaches of the Yellow River, in the spring and gathered with waterfowl migrating from the middle and west China, or even central Asia, resulting in virus transmission and evolution. The results provide important evidence regarding virus transmission along the migratory route, which connects the poultry culture zone in eastern China to the northern breeding ground.
Wuliangsuhai Lake and the Ordos Lakes in Inner Mongolia were identified as important stopovers for migrating whooper swans. In May 2015, wild ducks were infected with H5N1 at both of these locations. It is believed that the presence of the virus at the wetlands enabled infection to other waterfowl during the migration period 4,17 and that the whooper swans increased the potential risk of spreading the virus during their migration 5 . At whooper swan breeding sites in central Mongolia, H5N1 was isolated from individuals between 2005 and 2010. These breeding sites, such as Khunt Lake, Erkhel Lake, and Doityn Tsagaan Lake, are located at the breeding sites of the spring migration. Satellite data showed that the swans moved extensively around these lakes and rivers, which may have increased the risk of virus dispersal. The wetlands, which provide habitat for many waterfowl, are important factors that affect the spread of the virus. Therefore, the wetlands identified as key molting or breeding sites should be used as the main AIV surveillance sites. Prosser et al. confirmed that bar-headed geese, ruddy shelducks and other waterfowl crossing Qinghai Lake migrate to breeding sites in central Mongolia along the Central Asia Flyway 25 . According to the tracking data in our study, the same breeding sites were also selected by whooper swans. Therefore, there is an increased risk of AIV transmission between Qinghai Lake in western China and Sanmenxia in central China 28 . A similar increased risk appears to exist for other regions.
Through the regression analysis of the H5N1 events and the home range of whooper swan, we can see that avian influenza outbreaks are closely related to bird activity. Wild birds might stop for water and food at some key stopover sites that include the upper reaches of the Yellow River, Bayannur and Ordos in Inner Mongolia, and Yulin in Shaanxi Province along their migration flyways, promoting virus transmission among birds 29,30 . Thus, the denser the bird activity is, the higher the risk of bird flu outbreaks. Studying the home range is very important to preventing the outbreak of avian influenza.
In this study, a strain of H5N1 2.3.2.1.c subtype AIV was isolated from Yulin in Shaanxi, which is key stopover site for whooper swans during their spring migration. The results of the phylogenetic analysis showed that the strain has a high degree of homology (99.7%) with the epidemic strain that we previously isolated from the Sanmenxia Reservoir area. We confirmed that whooper swans could carry the HPAI virus during migration, resulting in long-distance transmission.
In May 2015, at the period of spring migration for whooper swans, an H5N1 outbreak occurred at important stopovers of whooper swans in Inner Mongolia. Approximately one hundred black-necked grebes (Podiceps nigricollis) were found dead at Tsagaan Nur of Ordos 31 (Fig. 2), and the H5N1 strain was closely related to sequences isolated from Sanmenxia, which further supports our conclusion.

Methods
Capture and marking. The Sanmenxia Reservoir area is located in the middle reaches of the Yellow River, central China (Fig. 1). In January and February from 2013 to 2016, a large cage with food such as corn was established in the Sanmenxia Swan Lake to attract whooper swans. During the 3 years of the study, a total of 61 swans were captured with the approval of the State Forestry Administration of China. All swans were marked with metal rings and neck collars (white "A/G+number" on blue, for example, A57), and the 61 healthiest swans received satellite transmitters attached to their back with a harness (Table 3). We recorded body weight, age and sex for each swan. Age was classified as adult or juvenile based on visual observation. All swans were released within 30 minutes near the capture sites after processing. The 61 marked swans were denoted by the letters S, from S1 to S61 (Table 3). History of H5N1 outbreak information. Detailed H5N1 outbreak data, including longitude and latitude, date, species, and virus subtypes, were obtained from the United Nations Food and Agriculture Organization (http://www.fao.org) and the World Organization for Animal Health (http://www.oie.int).

Satellite telemetry information.
Sample collection and phylogenetics. Tracheal and cloacal swab samples were taken from the 61 captured whooper swans for laboratory testing. Field work was also carried out in March and April 2015 to collect fecal samples along the routes of the satellite tracked swans. All samples were stored at −80 °C and inoculated into 10-day-old embrocated specific-pathogen free (SPF) chicken eggs for virus isolation at the National Avian Influenza Reference Laboratory of China.
Hemagglutinin (HA) gene sequences of HPAI H5N1 viruses were obtained from the GenBank hosted by the National Center for Biotechnology Information (NCBI) as of 10 November 2016. Sequences were aligned with MAFFT v7.127b and trimmed to a length of 1716 bp. An approximate maximum likelihood tree was constructed by FastTree v2.1.4 32 as the setting of GTR + Gamma with SH-like local-bootstrap support values to estimate the reliability of each split in the tree. The tree was rooted at A/goose/Guangdong/1/96 and structured according to the unified nomenclature system and virus clades for the HPAI H5N1 viruses of the World Health Organization.

Spatial analysis.
A stopover site was defined as an area where whooper swans stayed for a 72-h period or longer. A breeding site, breeding or/and molting site, indicated the northern limit of the spring migration route where swans stayed for more than 3 months. We classified the arrival date at the breeding site as the end of the spring migration for the swans. Hence, we only selected the GPS locations before the end of the spring migration for analysis. We classified swans that stayed in China in May and stopped migrating north as summer residents. Because the GPS signal was not always continuous, we defined the first quarter of the total period as the departure time and the last quarter as the arrival time, with the remainder as the flight or stopover time if the GPS signal was not continuous for a period of <5 consecutive days. Because the Yellow River in inner Mongolia as an important stopover site and ecological corridor where most of the whooper swans stayed for the longest time, we defined it to be migration critical and divided migration routes into 2 stages: (1) from wintering grounds to the Yellow River stretch in inner Mongolia, and (2) from the Yellow River stretch in inner Mongolia to breeding sites in Mongolia.
We ran the Brownian bridge movement model using the 'BBMM' package in Program R (version 3.4, R Development Core Team 2016) to estimate the population level of the 61 whooper swans in our data with requested contour levels of 95% and 99% 33,34 . The BBMM provides an empirical estimate of a movement path SCIENTIFIC REPoRTS | (2018) 8:7049 | DOI:10.1038/s41598-018-25291-1 using discrete location data obtained at relatively short time intervals, which is the probability of being in an area during the time of observation that is conditioned on starting and ending locations. We calculated all whooper swan output grids at the same spatial extent and at a 200 × 1000 km 2 grid resolution. We plotted and analyzed migration routes, time, and days at stopovers using ArcGIS version 10.2 (ESRI, Redlands, California, USA).
We set up 80 random points in the study area, and the ratio of random points to H5N1 events is 10 to 1. We used spatial analysis techniques for H5N1 events and random points in 95%, 99% and 100% of the home range by logistic regression in SPSS 20.0. Ethics statement. Approval