Abstract
The zoonotic origin of the SARS-CoV-2 pandemic is still unknown. Animal experiments have shown that non-human primates, cats, ferrets, hamsters, rabbits and bats can be infected by SARS-CoV-2. In addition, SARS-CoV-2 RNA has been detected in felids, mink and dogs in the field. Here, we describe an in-depth investigation of outbreaks on 16 mink farms and humans living or working on these farms, using whole genome sequencing. We conclude that the virus was initially introduced from humans and has evolved, most likely reflecting widespread circulation among mink in the beginning of the infection period several weeks prior to detection. At the moment, despite enhanced biosecurity, early warning surveillance and immediate culling of infected farms, there is ongoing transmission between mink farms with three big transmission clusters with unknown modes of transmission. We also describe the first animal to human transmissions of SARS-CoV-2 in mink farms.
One sentence summary SARS-CoV-2 transmission on mink farms.
Main text
Late December 2019, SARS-CoV-2 was identified as the causative agent in a viral pneumonia outbreak, possibly related to a seafood and a live animal market in Wuhan, China (1). Since then, SARS-CoV-2 spread across the world and by August 31rd, over 25,200,000 people had been infected with SARS-CoV-2 resulting in over 840,000 deaths (2). In the Netherlands, over 72,000 infections have been confirmed, over 6,200 SARS-CoV-2 related deaths have been reported, and drastic measures have been put into place to prevent further spread of SARS-CoV-2 (3).
In view of the similarities with SARS-CoV-1, a zoonotic origin of the outbreak was suspected by the possible link with the Wuhan market where various animals were sold including fish, shellfish, poultry, wild birds and exotic animals. The finding of cases with onset of illness well before the period observed in the cluster, however, suggests the possibility of other sources (4). Although closely related coronaviruses in bats (5, 6) and pangolins (7, 8) have most sequence identity to SARS-CoV-2, the most likely diversion date of SARS-CoV-2 from the most closely related bat sequence is estimated to date back to somewhere between 1948-1982 (9). Therefore, the animal reservoir(s) of SARS-CoV-2 is (are) yet to be identified.
Experimental infections in dogs (10), cats (10, 11), ferrets (10, 12), hamsters (13, 14), rhesus macaques (15), tree shrew (16), cynomolgus macaques (17), grivets (18), common marmosets (19), rabbits (20), and fruit bats (21) have shown that these species are susceptible to SARS-CoV-2, and experimentally infected cats, tree shrews, hamsters and ferrets could transmit the virus. In contrast, experimental infection of pigs and several poultry species with SARS-CoV-2 proved to be unsuccessful (10, 21, 22). SARS-CoV-2 has also sporadically been identified in naturally infected animals. In the USA and in Hong Kong, SARS-CoV-2 RNA has been detected in dogs (23). In the Netherlands, France, Hong Kong, Belgium and the USA, cats have tested positive by RT-PCR for SARS-CoV-2 (24–27). Furthermore, SARS-CoV-2 has been detected in four tigers and three lions in a zoo in New York (28). In Italy, the Netherlands and in Wuhan, antibodies to SARS-CoV-2 have been detected in cats (29–31). Recently, detected SARS-CoV-2 was detected in farmed mink (Neovison vison) that showed signs of respiratory disease and increased mortality (29, 32).
Thereafter, the Dutch national response system for zoonotic diseases was activated, and it was concluded that the public health risk of animal infection with SARS-CoV-2 was low, but that there was a need for increased awareness of possible involvement of animals in the COVID-19 epidemic. Therefore, from May 20th 2020 onwards, mink farmers, veterinarians and laboratories were obliged to report symptoms in mink (family Mustelidae) to the Netherlands Food and Consumer Product Safety Authority (NFCPSA) and an extensive surveillance system was set up (33).
Whole genome sequencing (WGS) can be used to monitor the emergence and spread of pathogens (34–37). As part of the surveillance effort in the Netherlands over 1,750 SARS-CoV-2 viruses have been sequenced to date from patients from different parts of the Netherlands (38). Here, we describe an in-depth investigation into the SARS-CoV-2 outbreak in mink farms and mink farm employees in the Netherlands, combining epidemiological information, surveillance data and WGS on the human-animal interface.
Methods
Outbreak investigation
Following initial detection of SARS-CoV-2 in mink on two farms on April 23rd and April 25th, respectively, as part of routine health monitoring done by the Royal GD Animal Health service and subsequent investigation by Wageningen Bioveterinary Research (WBVR), the national reference laboratory for notifiable animal diseases, a One Health outbreak investigation team was convened (39, 40). Subsequently, respiratory signs and increased mortality in mink was made notifiable by the Dutch Ministry of Agriculture, Nature and Food Quality and the farms were quarantined (no movements of animals and manure and visitor restrictions). On May 7th two other mink farms in the same region were confirmed to be infected. By the end of May the Dutch minister of Agriculture decided that all mink on SARS-CoV-2 infected farms had to be culled. Moreover, as the clinical manifestation of the infection was highly variable within and between farms, including asymptomatic infections, weekly testing of dead animals for SARS-CoV-2 infections became compulsory for all mink farms in the Netherlands. Moreover, a nation-wide transport ban of mink and mink manure, and a strict hygiene and visitor protocol was implemented. The first infected mink farms were culled from June 6th onwards. From the 10th infected farm (NB10) onwards, culling took place within 1-3 days after diagnosis. In this manuscript, the data up to June 26th, when a total of 16 mink farms in the Netherlands were found positive for SARS-CoV-2 infections, is presented.
Veterinary and human contact tracing
The Netherlands Food and Consumer Product Safety Authority (NVWA) traced animal related contacts with other mink farms. Backward and forward tracing of possible high-risk contacts was done in the framework of the standard epidemiological investigation by the NVWA (i.e. focused on movement of vehicles, visitors such as veterinary practitioners, (temporary) workers, sharing of equipment between farms and transport and delivery of materials, such as feed, pelts, carcasses and manure). Persons with possible exposure from this investigation, as well as farm owners and resident farm workers were asked to report health complaints to the municipal health service for testing and – in the case of confirmed infections – for health advice and further contact tracing. Farm owners and workers on infected mink farms were informed of potential risks and were given advice on the importance and use of personal protective equipment and hygiene when handling animals (41). The contact structure on the farms was assessed through in-depth interviews, to identify additional persons with possible exposure to mink. In order to provide an enhanced set of reference genome sequences, anonymized samples from patients that had been diagnosed with COVID-19 in the area of the same four-digits postal codes as farms NB1-NB4 in March and April 2020 were retrieved from clinical laboratories in the region.
SARS-CoV-2 diagnostics and sequencing
The presence of viral RNA in mink samples was determined using a RT-PCR targeting the E gene as previously described (42). For the human samples, diagnostic RT-PCR was performed for the E and the RdRp gene (42). In addition, serology was performed, using the Wantai Ig total and IgM ELISA, following the manufacturer’s instructions(43). For all samples with a Ct value <32, sequencing was performed using a SARS-CoV-2 specific multiplex PCR for Nanopore sequencing, as previously described (3). The libraries were generated using the native barcode kits from Nanopore (EXP-NBD104 and EXP-NBD114 and SQK-LSK109) and sequenced on a R9.4 flow cell multiplexing 24 samples per sequence run. Flow cells were washed and reused until less than 800 pores were active. The resulting raw sequence data was demultiplexed using Porechop (https://github.com/rrwick/Porechop). Primers were trimmed after which a reference-based alignment was performed. The consensus genome was extracted and positions with a coverage <30 were replaced with an “N” as described previously (44). Mutations in the genome compared to the GISAID sequence EPI_ISL_412973 were confirmed by manually checking the mapped reads and homopolymeric regions were manually checked and resolved by consulting reference genomes. The average SNP difference was determined using snp-dists (https://github.com/tseemann/snp-dists). All sequences generated in this study are available on GISAID.
Phylogenetic analysis
All available near full-length Dutch SARS-CoV-2 genomes available on 1st of July were selected (n=1,775) and aligned with the sequences from this study using MUSCLE (45). Sequences with >10% “Ns” were excluded. The alignment was manually checked for discrepancies after which IQ-TREE (46) was used to perform a maximum likelihood phylogenetic analysis under the GTR+F+I +G4 model as best predicted model using the ultrafast bootstrap option with 1,000 replicates. The phylogenetic trees were visualized in Figtree (http://tree.bio.ed.ac.uk/software/figtree/). For clarity reasons all bootstrap values below 80 were removed. To look at potential relationships with migrant workers, also all Polish sequences from GISAID (47) were included in the alignment (Supplementary table 1).
Mapping specific mutation patterns on mink farms and in mink farm employees
Amino acid coordinates are described in relation to the Genbank NC_045512.2 reference genome. Open reading frames were extracted from the genome alignment using the genome annotation as supplied with the reference genome. A custom R script was used to distinguish synonymous from non-synonymous mutations and non-synonymous mutations were visualized using a tile map from the ggplot2 package (48).
Geographical overview of mink farms in the Netherlands and SARS-CoV-2 positive farms
To protect confidentiality, SARS-CoV-2 positive mink farms were aggregated at municipality level. The datasets “Landbouw; gewassen, dieren en grondgebruik naar gemeente” and “Wijk-en Buurtkaart 2019” from Statistics Netherlands (CBS) were used (49). Maps were created using R packages sp (50), raster (51) and rgdal (52) and ArcGIS 10.6 software by ESRI.
Results
SARS-CoV-2 was first diagnosed on two mink farms in the Netherlands on April 23rd (NB1) and April 25th (NB2), respectively. After the initial detection of SARS-CoV-2 on these farms an in-depth investigation was started to look for potential transmission routes and to perform an environmental and occupational risk assessment. Here, we describe the results of the outbreak investigation of the first 16 SARS-CoV-2 infected mink farms by combining SARS-CoV-2 diagnostics, WGS and in-depth interviews.
Screening of farm workers and contacts
Farm owners of the 16 SARS-CoV-2 positive mink farms were contacted by the municipal health services to conduct contact investigation and samples were taken for RT-PCR-based and serological SARS-CoV-2 diagnostics. In total, 97 individuals were tested by either serological assays and/or RT-PCR. In total, 43 out of 88 (49%) upper-respiratory tract samples tested positive by RT-PCR while 38 out of 75 (51%) serum samples tested positive for SARS-CoV-2 specific antibodies. In total, 66 of 97 (67%) of the persons tested had evidence for SARS-CoV-2 infection (table 1).
Anthropozoonotic transmission of SARS-CoV-2
During the interview on April 28th, four out of five employees from NB1 reported that they had experienced respiratory symptoms before the outbreak was detected in minks, but none of them had been tested for SARS-CoV-2. The first day of symptoms of people working on NB1 ranged from April 1st to May 9th. For 16 of the mink, sampled on April 28th, and one farm employee, sampled on May 4th, a WGS was obtained (hCov-19/Netherlands/NoordBrabant_177/2020). The human sequence clusters within the mink sequences although it had 7 nucleotides difference with the closest mink sequence (Figure 1 and cluster A in figures 2 and 3). On farm NB2, SARS-CoV-2 was diagnosed on April 25th. Retrospective analysis showed that one employee from NB2 had been hospitalized with SARS-CoV-2 on March 31st. All samples from the 8 employees taken on April 30th were negative by RT-PCR but tested positive for SARS-CoV-2 antibodies. The virus sequence obtained from animals was distinct from that of farm NB1, indicating a separate introduction (Figure 2 and 3, cluster B).
Zoonotic transmission of SARS-CoV-2
On mink farm NB3 SARS-CoV-2 infection was diagnosed on May 7th. Initially all seven employees tested negative for SARS-CoV-2, but when retested between May 19th and May 26th after developing COVID-19 related symptoms, 5 out of 7 individuals working or living on the farm tested positive for SARS-CoV-2 RNA. WGS were obtained from these five individuals and the clustering of these sequences with the sequences derived from mink from NB3, together with initial negative test result and the start of the symptoms, indicate that the employees were infected with SARS-CoV-2 after the mink on the farm got infected. Also, an additional infection was observed based on contact-tracing: a close contact of one of the employees – who did not visit the farm – got infected with the SARS-CoV-2 strain found on NB3. Animal and human sequences from farm NB3 were related to those from farm NB1, but were both part of cluster A.
Similarly, on mink farm NB7 zoonotic transmission from mink to human most likely occurred. On this farm, SARS-CoV-2 infection in mink was diagnosed on May 31st and employees initially tested negative for SARS-CoV-2 but started to develop symptoms at a later stage. Samples were taken between June 10th and July 1st from 10 employees of which 8 tested positive for SARS-CoV-2 RNA. From 2 samples WGS could be generated from the employees which clustered together with the sequences from the animals from this farm.
Comparison with national reference database and enhanced regional sampling
The sequences generated from mink farms and from mink farm employees were compared with the national database consisting of around 1,775 WGS. In addition, to discriminate between locally acquired infections and mink farm related SARS-CoV-2 infection, and to determine the potential risk for people living close to mink farms, WGS was also performed on 34 SARS-CoV-2 positive samples from individuals who live in the same four-digit postal code area compared to the first four mink farms. These local sequences reflected the general diversity seen in the Netherlands and were not related to the clusters of mink sequences found on the mink farms, thereby also giving no indication of spill-over to people living in close proximity to mink farms (sequences shown in magenta, Figure 2). The sequences from the mink farm investigation were also compared to sequences from Poland (n=65), since many of the mink farm workers were seasonal migrants from Poland, but the these were not related.
Mink farm related sequence clusters
Phylogenetic analysis of the mink SARS-CoV-2 genomes showed that mink sequences of 16 farms grouped into 5 different clusters (Figure 2 and 3). Viruses from farms NB1, NB3, NB4, NB8, NB12, NB13 and NB16 belonged to cluster A, sequences from NB2 were a separate cluster (B), those from farms NB6, NB7, NB9 and NB14 grouped together in cluster C, NB5, NB8, NB10 and NB15 grouped to cluster D, and NB11 had sequences designated as cluster E. On farm NB8, SARS-CoV-2 viruses could be found from both cluster A and cluster D. A detailed inventory of possible common characteristics, like farm owner, shared personnel, feed supplier and veterinary service provider, was made. In some cases, a link was observed with the same owners of several farms, for instance for cluster A for NB1 and NB4, and for NB8 and NB12. Although NB7, NB11 and NB15 were also linked to the same owner, viruses from these farms belonged to cluster C, D and E respectively. No common factor could be identified for most farms and clustering could also not be explained by geographic distances as multiple clusters were detected in different farms located close to each other (Table 2 and figure 4).
In total 18 sequences from mink farm employees or close contacts were generated from seven different farms. In most cases, these human sequences were near-identical to the mink sequences from the same farm. For NB1 the situation was different and the human sequence clusters deeply within the sequences derived from mink (Figure 1), with 7 nucleotides difference with the closest related mink sequence. This was also the case on farm NB14, with 4 nucleotides difference with the closest related mink sequence. Employees sampled at mink farm NB8 clustered with animals from NB12 which can be explained by the exchange of personnel between these two farms.
Within farm diversity
SARS-CoV-2 was detected on mink farm NB1-NB4 after reports of respiratory symptoms and increased mortality in mink. The sequences from farm NB1 had between 0 and 9 nucleotides differences (average 3.9 nucleotides) and from NB2 between 0 and 8 nucleotides differences (average of 3.6), which is much more than what has been observed in outbreaks in human settings. The sequences from NB3 had 0 to 2 nucleotides difference suggesting that the virus was recently introduced, in line with the observed disease in humans, which occurred in the weeks post diagnosis of the infection in mink. After the initial detection of SARS-CoV-2 on mink farms, farms were screened weekly. The first, second, fifth and sixth weekly screening yielded new positives. The sequences of mink at NB6 had between 0 and 12 nucleotides differences, whereas diversity was lower for the subsequent farm sequences (Table 2).
Several non-synonymous mutations were identified among the mink sequences compared to the Wuhan reference sequence NC_045512.2. However, no particular amino acid substitutions were found in all mink samples (Figure 5). Of note, three of the clusters had the position 614G variant (clusters A, C and E), and 2 had the original variant. There were no obvious differences in the presentation of disease in animals or humans between clusters based on the data available at this stage, but further data collection and analysis, also for cases after NB16, are ongoing to investigate this further. The observed mutations can also be found in the general population and the same mutations also were found in human cases which were related to the mink farms.
Discussion
Here we show ongoing SARS-CoV-2 transmission in mink farms and spill-over events to humans. To the best of our knowledge, these are the first animal to human SARS-CoV-2 transmission events documented. More research in minks and other mustelid species, to demonstrate if these species can be a true reservoir of SARS-CoV-2 although from our observations we consider this likely. After the detection of SARS-CoV-2 on mink farms, 68% of the tested farm workers and/or relatives or contacts were shown to be infected with SARS-CoV-2, indicating that contact with SARS-CoV-2 infected mink is a risk factor for contracting COVID-19.
A high diversity in the sequences from some mink farms was observed which most likely can be explained by many generations of infected animals before an increase in mortality was observed. The current estimates are that the substitution rate of SARS-CoV-2 is around 1.16*10^-3 substitutions/site/year (53), which corresponds to around one mutation per two weeks. This could mean that the virus was already circulating in mink farms for some time. However, there was also a relatively high sequence diversity observed in farms which still tested negative one week prior, hinting towards a faster evolution of the virus in the mink population. This can indicate that the virus might replicate more efficiently in mink or might have acquired mutations which makes the virus more virulent. However, no specific mutations were found in all mink samples, making increased virulence less likely. In addition, mink farms have large populations of animals which could lead to very efficient virus transmission. Generation intervals for SARS-CoV-2 in humans have been estimated to be around 4-5 days(54), but with high dose exposure in a high-density farm could potentially be shorter. Recently, a specific mutation in the spike protein (D614G) was shown to result in an increased virulence in vitro (55), while it was not associated an increased growth rate for cluster nor an increased mortality (56). This mutation was present in farm clusters A, C and E, but no obvious differences in clinical presentation, disease severity, or rate of transmission to humans was observed.
While we found sequences matching with the animal sequences on several farms, not all of these can be considered direct zoonotic transmissions. For instance, the two employees from mink farm NB3 were most likely infected while working at the mink farm given the specific clustering in the phylogenetic tree and the timing of infection. Subsequent human infections may have originated from additional zoonotic infections, or from human to human transmission within their household. Further proof that animals were the most likely source of infection was provided by the clear phylogenetic separation between farm related human cases and animal cases, from sequences from cases within the same 4-digit postal code area. Spill-back into the community living in the same 4-digit postal code area was not observed using sequence data, but cannot be entirely ruled out as the testing strategy during April and May was focusing on health care workers, persons with more severe symptoms, and persons at risk for complications, rather than monitoring community transmission and milder cases.
While the number of SARS-CoV-2 infected individuals was decreasing in the Netherlands in May and June, an increase in detection of SARS-CoV-2 in mink farms was observed. Based on WGS these sequences are part of multiple individual transmission chains linked to the mink farms and are not a reflection of the situation in the human population during this time. In some cases, the farms had the same owner but in other cases no epidemiological link could be established. People coming to the different farms might be a source but also semi-wild cats roaming around the farms or wildlife might play a role (27). So far, the investigation failed to identify common factors that might explain farm to farm spread. During interviews, it became clear that farms had occasionally hired temporary workers that had not been included in the testing and were lost to follow-up, stressing the need for vigorous biosecurity and occupational health guidance. Since our observation, SARS-CoV-2 infections have also been described in mink farms in Denmark, Spain and the USA (57–59), and mink farming is common in other regions of the world as well, also in China where around 26 million mink pelts are produced on a yearly basis (60). The population size and the structure of mink farms is such that it is conceivable that SARS-CoV-2 – once introduced – could continue to circulate. Therefore, continued monitoring and cooperation between human and animal health services is crucial to prevent the animals serving as a reservoir for continued infection in humans.
Funding
This work has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 874735 (VEO), No. 848096 (SHARP JA) and No. 101003589 (RECoVER), from ZonMW (grant agreement No. 10150062010005), and from the Netherlands Ministry of Agriculture, Nature and Foods.
Authors contributions
Conceptualization: B.B.O.M, R.S.S, A.S., M.P.G.K; investigation: B.B.O.M, D.F.N., R.S.S, A.S., M.P.G.K, L.A.M.S, W.H.N.v.d.P, R.J.M, R.J.B, E.M, R.M, A.v.d.S, P.T, A.R, M.B, N.B-V, F.H, R.H.v.d.H, M.C.A.W-B, R.J.B, C.G, A.A.v.d.E, F.C.V, L.A.M.S; supervision: M.P.G.M; writing original draft: B.B.O.M, R.S.S, M.P.G.K; writing review and editing, all authors
Competing interests
Authors declare no competing interests.
Data and material availability
All data, code and materials used described in this manuscript are publicly available.