key: cord-0310499-ng1rdksu authors: Lu, Lu; Sikkema, Reina S.; Velkers, Francisca C.; Nieuwenhuijse, David F.; Fischer, Egil A.J.; Meijer, Paola A.; Bouwmeester-Vincken, Noortje; Rietveld, Ariene; Wegdam-Blans, Marjolijn C.A.; Tolsma, Paulien; Koppelman, Marco; Smit, Lidwien A.M.; Hakze-van der Honing, Renate W.; van der Poel, Wim H. M.; van der Spek, Arco N.; Spierenburg, Marcel A. H.; Molenaar, Robert Jan; de Rond, Jan; Augustijn, Marieke; Woolhouse, Mark; Stegeman, J. Arjan; Lycett, Samantha; Oude Munnink, Bas B.; Koopmans, Marion P. G. title: Adaptation, spread and transmission of SARS-CoV-2 in farmed minks and related humans in the Netherlands date: 2021-07-14 journal: bioRxiv DOI: 10.1101/2021.07.13.452160 sha: 28276982662ff4203788744f2ba665c7de60ec7c doc_id: 310499 cord_uid: ng1rdksu In the first wave of the COVID-19 pandemic (April 2020), SARS-CoV-2 was detected in farmed minks and genomic sequencing was performed on mink farms and farm personnel. Here, we describe the outbreak and use sequence data with Bayesian phylodynamic methods to explore SARS-CoV-2 transmission in minks and related humans on farms. High number of farm infections (68/126) in minks and farm related personnel (>50% of farms) were detected, with limited spread to the general human population. Three of five initial introductions of SARS-CoV-2 lead to subsequent spread between mink farms until November 2020. The largest cluster acquired a mutation in the receptor binding domain of the Spike protein (position 486), evolved faster and spread more widely and longer. Movement of people and distance between farms were statistically significant predictors of virus dispersal between farms. Our study provides novel insights into SARS-CoV-2 transmission between mink farms and highlights the importance of combing genetic information with epidemiological information at the animal-human interface. Since the initial cluster of cases reported in Wuhan, China, SARS-CoV-2 is predominantly 21 transmitted between people, with occasional examples of transmission between humans and 22 animals. An expanding range of animals has been found to be susceptible and natural infections 23 have been documented particularly in carnivores, including dogs, cats, lions and tigers, otters 24 and ferrets, which were in contact with infected humans 1,2 . Infections have not been detected 25 in most common livestock species, but multiple countries have reported SARS-CoV-2 in 26 farmed minks to the World Organisation for Animal Health (OIE) 27 (https://wahis.oie.int/#/dashboards/country-or-disease-dashboard). 28 29 In the Netherlands, SARS-CoV-2 was first detected in farmed minks in late April with signs 30 of respiratory symptoms and increased mortality 3 . An in-depth One Health investigation, 31 combining whole genome sequencing (WGS) with epidemiological information, was 32 conducted in response to the outbreaks in mink farms. The findings of the initial investigation 33 between April and June highlighted that mink sequences from the first 16 farms grouped into 34 8 The last infected farm was detected on the 4 th of November, after which no new infections were 128 detected ( Figure 1a) . 129 130 Spill-over into local community and limited onward transmission In total, 218 sequences isolated from randomly selected patients from 31 postal codes, in the 132 region of SARS-CoV-2 positive mink farms were obtained in period 4 th March 2020 to 4th 133 January 2021, to assess possible spill-over to the local community. In addition, all sequences 134 submitted to GISAID from the Netherlands until the 4 th of January were included in the analysis. 135 In three separate occasions, a mink related strain, linked to clusters A and C (Figure 1c) , was 136 detected. Two out of three patients infected with a mink strain (sampling dates in July and 137 August), lived in a province where no infected minks were reported, and they did not have 138 direct or indirect contact with the mink farming sector. One patient was found in the regional 139 screening in November but also did not report any mink farm contacts. After November, no 140 human infections with mink strains have been detected (Figure 1c) . 141 Throat swabs of the two escaped mink, caught 8 and 9 days in close proximity to two culled 142 farms (NB58 and NB59), at 450 and 650 m distance respectively, tested positive for SARS-143 CoV-2 RNA. Genome sequencing was successful for one mink sample, and it belonged to 144 Cluster A (Figure 1c) . 145 146 Specific mutations in the Spike protein in multiple mink clusters 147 We further explored how the specific mutations in the spike protein are associated with 148 phylogenies by mapping 4 potential important mutations in the spike protein (L452M, Y453F, 149 F486L, N501T) on the tree composed of the complete dataset ( Figure 2 ). These 4 mutations 150 9 are in confirmed contact residues of the viral spike protein with the ACE2 receptor 6, 7 . Within 151 the Netherlands, these 'mink specific' mutations were only found in minks and employees on 152 mink farms by the time the analysis has been performed (by 1 st April 2021), except for 3 153 samples: two sequences from unrelated humans (one with F486L, the other with both F486L 154 and L452M, third sequence excluded due to insufficient coverage) and one sequence from an 155 escaped mink (with F486L). However, these mutations have also been seen elsewhere in other 156 independent lineages. For example, the F486L has been detected occasionally in humans in 157 Ireland and Columbia, and in mink samples from the US (http://cov-glue.cvr.gla.ac.uk/#/home). 158 The 4 mutations have evolved in multiple clusters and in both human and mink samples from 160 Dutch mink farms. Specifically, mutation F486L has been seen in 217 sequences from 40 mink 161 farms that belong to 2 separate clusters (A and C), which accounted for 67% sequences and 162 68% sequences isolated within the cluster. Y453F has been seen in 37 sequences from 10 163 different farms in 3 different Clusters (A, D and E), which accounted for 3%, 82% and 100% 164 sequences isolated within the cluster. In addition, we found the N501T mutation in only 3 mink 165 virus sequences from 3 different farms belonging to Clusters A and D. L452M was seen in 44 166 sequences isolated from 9 mink farms all belonging to Cluster C (59%). N501T only appeared 167 in a short period of the outbreak (end of April to end of May), while the others appeared in a 168 later stage and sustained longer (F486L first appeared in two sequences in Cluster A at the end 169 of April, then reappeared and replaced F486 in Cluster A since mid-August and in Cluster C 170 since June, respectively); L452M appeared from early July to September and Y453F appeared 171 from end of April to early July. 172 We mapped 4 types of traits (host, farm ID, province and municipality) on individual time-174 scaled phylogenies of Cluster A, C and D using discrete trait models. We compared the 4 175 individual mutations in the spike protein and the combinations of the 4 mutations on the time-176 scaled phylogenies of Cluster A, C and D independently. The discrete trait mapping trees of 177 Cluster A are shown in Figure 3 , with the branches and nodes colored by inferred ancestral 178 traits. The trees for Cluster C and Cluster D are shown in Figure S1 and Figure S2 . The 179 occurrence of the mutations did not show any significant association either to host types, to 180 farm numbers or to locations (Mann-Whitney U Test, with p>0.5). 181 182 We compared the phylodynamics of three clusters (A, C and D). The results of estimating the 184 time to the most common recent ancestor (TMRCA), the molecular clock evolutionary rate and 185 spatial diffusion rate (geography.clock.rate) according to available data and parameters 186 selected are shown in Figure 4 . For Cluster A, the estimated TMRCA for mink sequences is 187 approximately in mid-March 2020 (mean 15 th March 2020 with 95% HPD (12 th March 2020, 188 28 th March 2020); Evolution rate is approximately 1.41 x10 -3 subst/site/year with 95% HPD 189 (1.2 x10 -3 , 1.75 x10 -3 ) subst/site/year. The other two clusters have slightly lower evolution rates 190 and more recent TMRCAs, but with wider HPD intervals; overall these results are consistent 191 with the estimations using a relax clock model on the complete data in Figure 1 . Similarly, the 192 spatial diffusion rate of Cluster A is higher (means of 2.91 x10 -4 ) than the other two clusters C 193 and D, which have means of 1.06 x10 -4 and 1.34x10 -4 ( Figure 4 and Table S1 ). Overall, the 194 TMRCA aligns with the epidemiological data about the emergence and detection of CoV-2 in the Netherlands. It also has a faster and wider spatial spread and higher evolutionary 196 rate than the other clusters. Figure 5 ). We observed similar results by 217 using the multi-type birth-death model which showed a strong increase in the number of 218 infections in clades with mutations rather than clades without mutations ( Figure S3 ). 219 Host (humans and minks) and farm number labels were added to the sequences, and the number 221 of transmissions between hosts (asymmetric) and between farms (symmetric) were inferred 222 using discrete traits models on the time resolved trees (Figure 3 , Figure S1 and S2). To avoid 223 sample size effect on the results, sequences were further subsampled to reduce over-224 representative sequences from the same farm. For transmissions identified by Markov jumps, 225 we also used BSSVS to identify only statistically significant pairs (with Bayes Factor >3, the 226 higher the value, the stronger the support). We summarised and compared the network among 227 three clusters A, C and D. 228 229 Overall, at least 43 zoonotic transmissions (with 95% HPD 34 to 50) from minks to humans 230 likely occurred in multiple farms (Table S2) some human infections may also be due to human-to-human infections, between mink farm 236 employees or farm owner family members, which is not included in the model. Therefore, the 237 true number of mink-to-human jumps may be lower. 238 There are also a few jumps between humans and minks from different farms. For example, 240 within Cluster A, a sequence from humans linked to NB49 are likely transmitted from minks 241 on NB47, although the low number of sequences (there is only one mink sequence obtained in 242 NB49) precludes robust conclusions. We found that viruses may jump back and forth between 243 humans and minks. The sequences sampled from humans in NB8 are likely transmitted to 244 minks in NB12, as shown in the phylogeny of Cluster A ( Figure 3) . Epidemiology data indeed 245 shows that the two farms have personnel links, which could be the explanation of this 246 observation (supporting file 1). 247 We also identified different potential transmission patterns networks between farms in Cluster 249 A, C and D ( Figure 6 and Table S3 ). For Cluster A, NB47 seems to be the most important 250 13 donor, with transmission to 7 farms (Figure 6a ). In comparison, fewer significant between farm 251 transmissions are identified in Cluster C and D (Figure 6b and c). Transmissions were also 252 drawn as links between locations of mink farms on the map ( Figure 7 ). Interestingly, we found 253 transmissions with high BF supports (darker red edges in Figure 7a ) that are not necessarily 254 between adjacent farms, and sequences from adjacent farms with personnel links (e.g., NB58 255 and NB59). In addition, sequences from different barns on the same farms do not necessarily 256 group together. For example, within Cluster C, sequences isolated from NB6 at the same date 257 fell into two separate sub-clades. 258 Assuming the presence of farm specific signatures allowed linking cases to farms, the two 260 unrelated human sequences are most closely related to sequences from to farms NB17 and 261 NB58, respectively; and the sequence from an escape mink is likely to have a relation with 262 farm NB65 ( Figure 6 ). However, the patients infected with mink strains did not report any 263 direct or indirect contact with mink or mink farm employees. 264 265 During our study, a detailed inventory of possible common characteristics, including farm 267 owner, shared personnel, feed supplier and veterinary service provider was made. 268 Epidemiological investigation indicated that many farms shared the same feed supplier or 269 veterinarian, but no unambiguous service company contacts were found between farms within 270 the different virus clusters which could explain the farm-to-farm spread. For 55% of the SARS-271 CoV-2 positive farms, owners, family members or personnel, including people with limited 272 contact with minks, were shared between farms (supporting file 1). 273 Using a generalized linear model (GLM), implemented in BEAST, we tested the contribution 275 of a range of predictor variables to the spread of viruses between farms which was estimated 276 in the discrete trait phylogeographic model ( Figure 6 ). Correlations between the predictor data 277 collected from mink farms were tested and highly correlated predictors were omitted ( Figure S4) . 278 The predictors being tested are 1) distance between farms; 2) personnel links between farms 3) 279 feed supplier; 4) veterinary service provider; 5) mink population per farm; 6) number of 280 sequences per farm included in the phylogenetic analysis; 7) human population density in 281 municipality where farm was located; 8) days between sampling and culling per farm 282 (supporting file 1). 283 For Cluster A, the distance between farms had a negative impact on the transmission between 284 farms (Table 1) , which indicated that farms that are further apart have generally lower rate of 285 transmission between them; while farms with personnel links have a positive impact on the 286 transmission between farms, which could be an explanation of the strong supported long-287 distance diffusion observed in Figure 7 . For Cluster C and D, none of the predictors have 288 significant impact on the overall transmission between farms (Table S4) . be strongly associated with the mink infections and may be associated with decreased antibody 318 binding and increased ACE2 affinity [9] [10] [11] . Cluster V viruses were also found to infect humans 319 and were associated with community transmission after mink-to-human transmission 12 . The mean TMRCA and 95%HPD interval for each cluster. b The mean clock rate and 95% HPD interval for each Cluster. c The mean spatial diffusion rate and 95% HPD for each cluster. J a n − 0 1 J a n − 1 5 J a n − 0 1 J a n − 1 5 Host J a n − 0 1 J a n − 1 5 N T N501T J a n − 0 1 J a n − NB1 NB12 NB13 NB16 NB20 NB21 NB27 NB3 NB33 NB35 NB36 NB38 NB4 NB40 NB42 NB43 NB44 NB46 NB47 NB48 NB49 NB50 NB51 NB52 NB53 NB54 NB55 NB56 NB57 NB58 NB59 NB60 NB61 NB62 NB63 NB64 NB65 NB67 NB68 NB8 notfarm The likelihood of exposure of humans or animals to SARS-CoV-2 in COVID-19 SARS-CoV-2 Seroprevalence in Household Domestic Ferrets (Mustela putorius furo) SARS-CoV-2 infection in farmed minks, the Netherlands Transmission of SARS-CoV-2 on mink farms between humans and mink and back to humans A dynamic nomenclature proposal for SARS-CoV-2 lineages to assist genomic epidemiology Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 receptor Acquisition of the L452R mutation in the ACE2-binding interface of Spike protein triggers recent massive expansion of SARS-Cov-2 variants. bioRxiv Rapid SARS-CoV-2 whole-genome sequencing and analysis for informed public health decision-making in the Netherlands The SARS-CoV-2 Y453F mink variant displays a pronounced increase in ACE-2 affinity but does not challenge antibody neutralization Acquisition of the L452R mutation in the ACE2-binding interface of Spike protein triggers recent massive expansion of SARS-Cov-2 variants. bioRxiv An emerging SARS-CoV-2 mutant evading cellular immunity and increasing viral infectivity. bioRxiv SARS-CoV-2 Transmission between Mink (Neovison vison) and Humans SARS-CoV-2 and the human-animal interface: outbreaks on mink farms Estimated transmissibility and impact of SARS-CoV-2 lineage B.1.1.7 in England Monitoring of SARS-CoV-2 infection in mustelids Transmission risks and control of foot-and-mouth disease in The Netherlands: spatial patterns Local spread of classical swine fever upon virus introduction into The Netherlands: Mapping of areas at high risk Risk maps for the spread of highly pathogenic avian influenza in poultry SARS-CoV-2 in Danish Mink Farms: Course of the Epidemic and a Descriptive Analysis of the Outbreaks in 2020 SARS-CoV-2 Exposure in Escaped Mink First Description of SARS-CoV-2 Infection in Two Feral American Mink (Neovison vison) Caught in the Wild SARS-CoV-2 infection in cats and dogs in infected mink farms Occupational and environmental exposure to SARS-CoV-2 in and around infected mink farms. medRxiv Preliminary report of an outbreak of SARS-CoV-2 in mink and mink farmers associated with community spread The B1.351 and P.1 variants extend SARS-CoV-2 host range to mice. bioRxiv, 2021 Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR Minimap2: pairwise alignment for nucleotide sequences The Sequence Alignment/Map format and SAMtools MAFFT: iterative refinement and additional methods IQ-TREE 2: New Models and Efficient Methods for Phylogenetic Inference in the Genomic Era Exploring the temporal structure of heterochronous sequences using TempEst (formerly Path-O-Gen) Bayesian phylogenetic and phylodynamic data integration using BEAST 1.10 Bayesian estimation of past population dynamics in BEAST 1.10 using the Skygrid coalescent model Improving the Accuracy of Demographic and Molecular Clock Model Comparison While Accommodating Phylogenetic Uncertainty Relaxed phylogenetics and dating with confidence Birth-death skyline plot reveals temporal changes of epidemic spread in HIV and hepatitis C virus (HCV) BEAST 2.5: An advanced software platform for Bayesian evolutionary analysis Phylodynamics with Migration: A Computational Framework to Quantify Population Structure from Genomic Data Comparing SARS-CoV-2 with SARS-CoV and influenza pandemics Bayesian phylogeography finds its roots Learning to count: robust estimates for labeled distances between molecular sequences 1 We acknowledge the authors, originating and submitting laboratories of the sequences from 2 GISAID's EpiCov Database on which the phylogenetic analysis was based (see Supplement). 3All submitters of data may be contacted directly via the GISAID website www.gisaid.org. Quality. 9 10 We declare that we have no competing financial, professional, or personal interests that might 11 have influenced the performance or presentation of the work described in this article. All authors provided critical feedback and contributed to manuscript editing. 20 21 Competing Interests 22 The authors declare no competing interests 23