Genome-wide Investigation of Genetic Diversity in the Poultry Red Mite, Dermanyssus Gallinae, From European Farms Utilising a NGS-Multiplex Platform

The poultry red mite (Dermanyssus gallinae), an obligatory blood feeding ectoparasite, is primarily associated with egg laying hens where it is estimated to cause losses of ~€230 million per annum from European farmers. Current control strategies, including the use of acaricidal chemicals and desiccant dusts, are often ineffective and there is widespread resistance to acaricides across Europe. Alternative methods to control D. gallinae are urgently required and strategies include development of recombinant subunit vaccines and discovery of new potential acaricides. These strategies will benet hugely from knowledge of the extent and rates of occurrence of genetic diversity within D. gallinae populations. In this study, genetic diversity of mites harvested from the UK and from sites across mainland Europe was studied at inter- and intra-farm levels. To achieve this, the genome analysis toolkit (GATK) best practices pipeline for single nucleotide polymorphism (SNP) and insertion/deletion variant calling was modied to be self-validating and used to identify 32,599 D. gallinae SNPs by comparing transcriptomic sequences (derived from mites harvested in Germany, Schicht et al.) with a D. gallinae genome assembly (derived from mites harvested in Scotland, Burgess et al.). Dermanyssus gallinae populations were sampled from 22 UK farms and 57 farms from 15 countries in mainland Europe. Analysis of 144 high-quality SNP markers across 117 pooled D. gallinae samples showed high spatial genetic diversity with signicant linkage disequilibrium. Revisiting a subset of farms revealed notable temporal changes in genetic diversity. reductive involving select criteria: minimum read depth of 36 and a minimum PHRED quality score of 350. SNPs excluded if they were located within the rst or last 50bp of a genomic contig. To optimise genome coverage SNPs were selected from individual contigs, emphasis on individual SNPs incorporated from the largest contigs available in the genome assembly. An SNPs were incorporated into the panel based on location within a region anking one of the 100 targeted SNPs, given a minimum PHRED quality score of 250 and a minimum read depth of 20. using MEGA-X (Kumar et al., Statistical comparison achieved using 1,000 bootstrap iterations.


Introduction
Dermanyssus gallinae (de Geer) is an obligatory blood feeding ectoparasite (Chauve, 1998). Primarily a parasite of birds, most notably laying hens, D. gallinae demonstrates considerable plasticity regarding host speci city and is capable of feeding on mammals, including humans (Valiente Moro et al., 2009). Dermanyssus gallinae is reported to have a worldwide distribution with high percentages of laying hen farms affected in European countries including Denmark, France, Romania, Italy, the Netherlands, Poland, Serbia and the United Kingdom (UK) (Sparagano et al., 2014, Hoglund et al., 1995, Guy et al., 2004, Fiddes et al., 2005, Cencek, 2003. In the UK, for example, between 60% and 85% of commercial hen egg laying systems are reported to be infested (Guy et al., 2004, Fiddes et al., 2005. Dermanyssus gallinae causes signi cant economic losses to the European poultry industry, estimated at ~€231 million per annum (Van Emous, 2017) attributed to higher feed conversion ratios, reduced quality and number of eggs, and the cost of control (Sparagano et al., 2014). Losses in the UK alone were estimated at €3 million in 2008 (Sparagano et al., 2014) and infestation rates have increased signi cantly since then. Affected hens become restless and display signs of itching/irritation with severe infections causing anaemia which can, especially in young hens at point of lay, cause death (Marangi et al., 2009b). Research by Kilpinen et al. (2005) on the in uence of D. gallinae infestation on laying hen health showed a reduction in weight gain in young hens when compared to hens without infestation that persisted for at least 100 days (Kilpinen et al., 2005). In addition to the direct impacts of infestation, it is suggested that D. gallinae plays a role in bird-to-bird transmission of other pathogens including some that are zoonotic. For example, Newcastle disease virus has been isolated from D. gallinae mites (Arzey, 1990) and it has been shown experimentally that D. gallinae is capable of transmitting Pasteurella multocida and Salmonella enterica enterica serovar Gallinarum between birds (Petrov, 1975, Cocciolo et al., 2020.
Control of D. gallinae most commonly relies on the use of various classes of chemical compounds collectively referred to as 'acaricides' (Sparagano et al., 2014), although widespread resistance to many current products has been demonstrated (Marangi et al., 2012, Katsavou et al., 2020, Marangi et al., 2009a. Dermanyssus gallinae infestation is increasingly common in Europe, its presence enhanced by bans on the use of some effective chemical treatments, as well as legislative changes that has seen traditional caged systems replaced by enriched cages whose more complex structures facilitate the survival and spread of mites ). Novel control methods are urgently required to reduce the health, welfare and economic impacts of D. gallinae and this includes screening for novel drugs as well as research into the development of recombinant vaccines. Vaccination appears to be a feasible approach for controlling D. gallinae (Bartley et al., 2012, Wright et al., 2016, Bartley et al., 2017, Bartley et al., 2015 but optimal antigens and strategies for delivery are yet to be determined. Given the rapidity with which D. gallinae populations can become resistant to many acaricides, it is not clear how mite populations would respond to being targeted by recombinant vaccines or the rapidity with which immune escape may occur. Improving knowledge of population structures for D. gallinae including the extent of naturally occurring genetic diversity and potential for transmission of resistance/escape alleles will provide data to underpin future models of genome evolution in response to novel control strategies. The importance of D. gallinae and opportunities offered by modern sequencing and genotyping platforms has seen recent increases in genetic studies and data in public databases. Studies of diversity have mainly focused on the cytochrome c oxidase subunit I (COI) coding sequence (Marangi et al., 2009b, Oines andBrannstrom, 2011, Chu et  . They reported intra-species variation rates of <9% in the COI gene and, in conjunction with further analysis, suggested that D. gallinae represents a complex of hybridized lineages, possibly species, from a total of 35 haplotypes (Roy et al., 2010). Research by Roy et al. has previously revealed that D. gallinae sensu lato represents a species complex, with a minimum of at least two cryptic species present (where a cryptic species can be de ned as one that cannot be distinguished just by morphological features). They de ne two cryptic species: D. gallinae sensu stricto and D. gallinae L1, with the species having been recorded in poultry farm populations across the world but not in other avians (Roy and Buronfosse, 2011, Roy et al., 2010, Roy et al., 2009a. In 2014, the rst next-generation sequence (NGS) dataset for D. gallinae was transcriptome data published (Schicht et al., 2014). Total RNA was extracted from an acaricide-susceptible D. gallinae strain maintained at the University of Veterinary Medicine Hannover, Institute for Parasitology. Synthesis of cDNA was completed using a pool of RNA extracted from male and female mites, including all developmental stages in starved and  (Burgess et al., 2018). They extracted genomic DNA from adult female D. gallinae mites and freshly laid eggs (collected from a layer farm in Scotland, UK), before using a combination of sequencings from PacBio and Oxford Nanopore Technologies MinION to produce a nal assembly of 7,171 contigs with an assembled genome size of 959 Mb (Burgess et al., 2018).
In this paper we used these published D. gallinae genomic and transcriptomic resources to identify a panel of highquality SNPs with utility for genome-wide population genetic analyses. Using a Mid-Plex SNP genotyping assay we have assessed the occurrence and extent of genetic diversity in UK and other European D. gallinae populations, de ning the occurrence of spatial and temporal variation.  S1). Mites were captured using cardboard traps as previously described (Nordenfors and Chirico, 2001). Thirty-one samples were collected from free range hen egg systems, of which 11 were organic, with seven collected from non-organic intensive enriched cage production systems. Three farms (three samples) provided D. gallinae mites but did not provide details relating to production system used or organic status. Nineteen farms were sampled once, contributing one sample from a single barn (n=11), two samples from different sections of a single barn (n=7), or two samples from different barns (n=1). Three farms were revisited on three (n=2) or four (n=1) occasions, contributing multiple samples (Supplementary Table S1). Mites were either used directly (fresh), dried and frozen at −20°C, or preserved in ethanol (>70% v/v; VWR International, Fontenary-sous-Bois, France). All farmers received instructions on how to carry out collection, a questionnaire and an information sheet relating to the project. Questions relating to production system and style, chicken husbandry and breed and farmer opinion regarding D. gallinae were included. Ethical approval was given by the Social Science Research Ethical Review Board at the Royal Veterinary College (URN SR2017-2018) for work pertaining to the questionnaire (supplied in Supplementary File 1). Con dential information was kept secure and all results were presented anonymously.

Mainland Europe
Dermanyssus gallinae collected from layer chicken farms in mainland Europe were received preserved in 70-100% (v/v) ethanol, or alive in cardboard traps which were either used directly, dried and frozen at −20°C, or preserved in ethanol (>70% v/v; VWR International, Fontenary-sous-Bois, France). Samples were received from 15 countries from a total of 57 farms ( Figure 1, Supplementary Table S2), with a mixture of samples collected and sent for this study and others sent from archives collected for previous research. Samples from farms were provided individually for all countries, except for Macedonia, where four farms were sampled and mites were pooled into a single tube.

Geographical clustering
Countries were grouped into four geographical clusters for analysis based on spatial proximity and climatic factors, with geographic cluster ID's assigned ( Table 1). The UK formed one cluster, recognising its physical separation from mainland Europe. . Due to a lack of preexisting known variants in D. gallinae, the GATK pipeline was modi ed for self-validation. Speci cally, one initial round of SNP discovery was undertaken with no variant recalibration step completed and the resulting VCF le (produced through intersection of the VCF les from both individual paired-end read sets) was used in the second round of SNP discovery to enable variant recalibration. Base quality score was recalibrated to produce quality scores and corrected for variation in quality based on sequence context and machine cycle. Variant calling was performed using "Uni edGenotyper" and "SelectVariants" arguments in GATK with the following ltering criteria to avoid possible false positives: SNPs with a Phred-scaled quality score of <30.

SNP selection
A panel of 100 high quality SNPs were identi ed based on a reductive system involving select criteria: minimum read depth of 36 and a minimum PHRED quality score of 350. SNPs were excluded if they were located within the rst or last 50bp of a genomic contig. To optimise genome coverage SNPs were selected from individual contigs, with emphasis placed on individual SNPs incorporated from the largest contigs available in the genome assembly. An additional 57 SNPs were incorporated into the panel based on location within a region anking one of the 100 targeted SNPs, given a minimum PHRED quality score of 250 and a minimum read depth of 20.

SNP analysis
Three criteria were applied to raw SNP genotyping data to permit inclusion in analysis; 1) a minimum read cut off of three, 2) a SNP call rate of 70% or higher across the assay panel, and 3) SNPs called from 90% of samples or higher were retained for use. Heterozygous calls were converted to the dominant allele using raw allele read depths generated during sequencing. To achieve this, the allelic fraction was calculated based on ALT/(REF+ALT), where REF = reference allele read depth and ALT = alternative allele read depth. A ratio of 0.6 or higher was used for stringent inference of alternative alleles and a ratio lower than 0.6 used to infer reference alleles.

Summary statistics from multiplex PCR SNP genotyping
One-hundred-and-fty-seven SNPs were analysed in two batches of 68 and 108 pooled D. gallinae samples (in plates 1 and 2), achieving 53.0% and 76.6% call rates, respectively (Tables 2-3). The SNP call rate was variable across plates, but in total 80.0% and 97.5% of SNPs achieved a call rate of greater than 50% per plate ( Table 2). SNP call rate was notably higher from plate 2, where 47.2% of samples were characterised at more than 80% of SNPs. Comparison of results by sample over both plates found that 89% achieved a call rate of 50% or greater (Table 3). After quality control a total of 117 samples with su cient genotype coverage were retained for analysis, permitting 144 SNPs to be assessed (Table  3). Overall, 131 of these SNP markers were found to be polymorphic. Thirteen markers were invariant, 12 of which represented reference alleles and one represented the alternative from a total of 117 pooled D. gallinae samples.

Nucleotide analysis
Mean genetic diversity for all samples combined was 0.3478 (calculated using LIAN; Table 4). Based on geographical proximity, cluster two had the lowest diversity (0.2915; Belgium, Czech Republic, Denmark, Germany and the Netherlands) whilst cluster four had the highest (0.3581; Italy, Croatia and Slovenia). Across all populations, no shared haplotypes were observed with a unique haplotype recorded for every sample (Table 4). Table 4 Linkage analysis of 145 SNP markers for 117 pooled D. gallinae samples using LIAN (84). Results shown for each dataset analysed including the full dataset, and subsets representing different production systems across Europe, four geographic clusters, UK production system and UK organic status. The number of samples and number of haplotypes included for each dataset was a shown. *Four farms were removed from production type analysis due to lack of information regarding production system. ** Total of 25

Production systems
Comparing production systems across Europe found no statistically signi cant difference in mean genetic diversity, with intensive production systems marginally higher. It should be noted that the UK accounted for most of the free-range farms sampled (77.5%) and conversely samples from UK intensive farms represented just 7/ 41 pools analysed and should be open to interpretation. In the UK, lower genetic diversity was observed in organic free-range samples ( Table 4).

Linkage disequilibrium
Linkage analysis utilising LIAN revealed signi cant linkage disequilibrium in the full dataset, all geographic clusters analysed, and all production systems (Table 4). Linkage disequilibrium (LD) analysis completed using DnaSP included 1653 pairwise comparisons and revealed a total of 483 signi cant pairwise comparisons by chi-square test, 74 of which remained signi cant using the Bonferroni procedure (average D' = 0.42927). D' (normalised D) was positive at all statistically signi cant sites, with and without the Bonferroni correction, indicating that these markers occurred together in the same haplotype more than expected.

Phylogenetic analysis and Network analysis
Exactly 117 haplotypes were identi ed through network analysis, corresponding to the 117 D. gallinae pooled samples analysed after quality control, with no shared haplotypes between samples ( Figure 2). Clustering of samples originating from the same country was observed for countries such as Belgium and Romania ( Figure 2). Conversely, haplotypes from more intensively sampled countries such as the UK were distributed across the network map, sharing phylogenetic relatedness with multiple countries. ML phylogenetic analysis revealed two main clades (A and B), with four subgroups  Figure 4A). Samples from UK farms were shown to cluster together (27/41 farms), in close relationship with farms from Albania, Greece and the Netherlands. In contrast, some individual haplotypes from UK were shown to be phylogenetically related to those from farms in Belgium, Portugal and Romania (Figure 2). Analysis of variation by production system in the UK demonstrated close phylogenetic relationships between all production systems and no clear differentiation between organic or non-organic free-range farms ( Figure 4B).
Haplotypes were assessed from samples collected from the same barn on farm UK6 on four occasions, starting with a single collection from one ock (time 0), followed by another three samples 14, 15 and 20 months later from a subsequent ock. While all four haplotypes were similar, those collected from the same ock were most closely related ( Figure 4A). A close phylogenetic relationship was observed between three samples collected from the same barn close together in time (six months apart; 14, 15 and 20 months after initial sampling, respectively), with greater phylogenetic separation seen from the initial (more distant) time of sampling ( Figure 4B). The ock was 1.3 years old at the rst time point (+0, Figure 4B), with a change in ock before the subsequent time points (+14, +15, +20), providing one plausible explanation for phylogenetic differentiation. Samples collected from two other barns on farm UK6 were phylogenetically distinct, with haplotypes from barns b and c more closely related to UK7 and UK18, respectively Samples from UK7 were found distributed across the network map, whilst samples from UK11 clustered close together. Two barns sampled from UK14 also demonstrated distinct haplotypes with clear phylogenetic differentiation between populations, with the haplotype from one barn in haplogroup A and the haplotype from the other barn found in haplogroup B ( Figure 4B).

Discussion
Knowledge of relevant population structure and pre-existing genetic diversity will be valuable for accurate prediction of longevity and e cacy of novel control methods against D. gallinae. To gain insights into D. gallinae population genetics, a custom Mid-Plex SNP genotyping assay was developed. 157 SNP markers were chosen from a panel of 30,199 high quality SNPs identi ed within the D. gallinae genome by comparison with transcriptomic data using an adapted selfvalidating GATK pipeline. A total of 144 SNP markers were analysed after three criteria were applied to raw SNP genotyping data, with 132 SNPs being informative (i.e. polymorphic). A distinct population structure was observed including high genetic diversity, de ned by multiple unique haplotypes, together with notable linkage disequilibrium.
No statistical difference in genetic diversity of D. gallinae populations from differing production systems was noted although non-organic farms in the UK had a consistent tendency to greater genetic diversity than organic farms. This could be due to strict limitations on use of chemicals in organic farms (Zenner et al., 2003) as Roy and Buronfosse have shown that selection imposed by control measures and hygiene practices affect allelic composition of D. gallinae populations (Roy and Buronfosse, 2011). Thus, greater pressure for diversifying selection exerted by the inclusion of chemical treatments alongside other methods may underpin increased genetic diversity in non-organic farms. Typically, selection pressure from the use of highly effective chemical control would be expected to reduce genetic diversity through imposition of a bottleneck (Coles and Dryden, 2014). However, personal observations during mite collection in the UK as well as questionnaire data revealed that for many farms a broad range of control measures were adopted, including chemical control, desiccant dusts, natural control measures and hygiene related measures (Supplementary  Table S3 Results from phylogenetic, network and linkage analyses suggest a spatial structuring of genetic diversity in D. gallinae across Europe, with high haplotype numbers for populations and signi cant linkage disequilibrium in all populations (P<0.01) ( Table 4). Admixture within UK farms (i.e. between Scotland, Ireland, Wales and England) was observed through close phylogeny and there were also close phylogenetic relationships between mites sampled from farms in different European countries. This is indicative of on-going admixture between farms in a single country and between countries, which has also been shown in studies of COI genetic diversity (Oines and Brannstrom . Spatial structuring of genetic diversity can occur when gene ow is insu cient for homogenisation of allele frequencies throughout a studied area (Bohonak, 1999). As well as gene ow and dispersal, genetic structure can be in uenced by genetic drift in small populations, which impacts genetic structure through increased differentiation (Levy and Neal, 1999). The sampling strategy for this study predominantly focused on commercial laying farms, where it is likely that the D. gallinae populations sampled repeatedly started from small founding populations including survivors from farm cleaning and/or new mites bought onto farms via infested hens or equipment (Roy et al., 2021). A subdivision in populations or changes in population size, as well as the exchange of individuals across populations, can all affect LD (Slatkin, 1975, Slatkin, 2008. Intentional or unintentional mixing of individuals from subpopulations with differing allele frequencies creates genomic LD (Mitton et al., 1973, Nei andLi, 1973). Mixing of D. gallinae populations during contraction and expansion can occur when infested transport/equipment or hens are bought onto farms and this can occur within individual countries from suppliers or across countries through trade links. This is supported by previous research focusing on Tropomyosin. Roy and Buronfosse demonstrated marked divergence between tropomyosin alleles carried by D. gallinae mites from farms compared to wild avifauna (Roy and Buronfosse, 2011). From poultry farms, two to three very distant tropomyosin alleles were commonly observed with heterozygous status, whilst in wild avifauna heterozygotes carried less distant alleles that could typically be assigned to one of three groups found in poultry farms (Roy and Buronfosse, 2011). The authors hypothesised that the major divergence of alleles observed in D. gallinae collected from farms could result from hybridisation events caused through the breakdown of geographical barriers by long-distance transfer through human involvement.
Network analysis revealed 117 unique haplotypes across all D. gallinae populations sampled with no shared haplotypes ( Figure 2). As genetic drift can impact genetic structure in small populations through increased differentiation (Levy and Neal, 1999), it is possible that small founding populations in farms have differentiated through genetic drift during population expansion and colonisation of individual farms, shifting allelic frequencies to form individual haplotypes. Investigation into COI diversity in Sweden and Norway revealed signs of differentiation between populations, consistent with the varied administrative processes in each country (Oines and Brannstrom, 2011). While the haplotypes detected were closely related, none were shared between the countries (Oines and Brannstrom, 2011). Sweden was also shown to share haplotypes with other European populations, possibly due to foreign trade and EU status (Oines and Brannstrom, 2011). Results demonstrated here regarding the genetic diversity of D. gallinae have implications for future control measures and could directly contribute to targeting novel control measures at highly conserved genes that encode proteins with an essential function. These genes are less likely to evolve diversity that promotes chemical resistance or vaccine escape due to selection pressure to maintain sequence conservation, making them optimal drug and vaccine candidates (

Conclusion
Analysis of 144 SNP markers from 117 pooled D. gallinae samples collected from across the UK and mainland Europe showed high spatial genetic diversity, with no conserved haplotypes detected at more than one location or on more than one occasion. Signi cant linkage disequilibrium was detected across all populations indicating historical and on-going admixture between D. gallinae populations.

Declarations
Ethical Approval and Consent to Participate Ethical approval was given by the Social Science Research Ethical Review Board at the Royal Veterinary College (URN SR2017-2018) for work pertaining to the questionnaire (supplied in Supplementary File 1). Con dential information was kept secure and all results were presented anonymously.

Competing Interests
The authors declare that the research was conducted in the absence of any commercial or nancial relationships that could be construed as a potential con ict of interest. Approximate location and characteristics of farms sampled for Dermanyssus gallinae across the United Kingdom and from 15 mainland European countries. Key provided outlines production system utilised by farm. One UK farm provided no address so has not been mapped directly. Free-range system encompasses free-range layer farms and backyard poultry ock.