Mobile Elements Harboring Heavy Metal and Bacitracin Resistance Genes Are Common among Listeria monocytogenes Strains Persisting on Dairy Farms

ABSTRACT Listeria monocytogenes is a foodborne pathogen and a resilient environmental saprophyte. Dairy farms are a reservoir of L. monocytogenes, and strains can persist on farms for years. Here, we sequenced the genomes of 250 L. monocytogenes isolates to investigate the persistence and mobile genetic elements (MGEs) of Listeria strains inhabiting dairy farms. We performed a single-nucleotide polymorphism (SNP)-based phylogenomic analysis to identify 14 monophyletic clades of L. monocytogenes persistent on the farms for ≥6 months. We found that prophages and other mobile genetic elements were, on average, more numerous among isolates in persistent than nonpersistent clades, and we demonstrated that resistance genes against bacitracin, arsenic, and cadmium were significantly more prevalent among isolates in persistent than nonpersistent clades. We identified a diversity of mobile elements among the 250 farm isolates, including three novel plasmids, three novel transposons, and a novel prophage harboring cadmium resistance genes. Several of the mobile elements we identified in Listeria were identical to the mobile elements of enterococci, which is indicative of recent transfer between these genera. Through a genome-wide association study, we discovered that three putative defense systems against invading prophages and plasmids were negatively associated with persistence on farms. Our findings suggest that mobile elements support the persistence of L. monocytogenes on dairy farms and that L. monocytogenes inhabiting the agroecosystem is a potential reservoir of mobile elements that may spread to the food industry. IMPORTANCE Animal-derived raw materials are an important source of L. monocytogenes in the food industry. Knowledge of the factors contributing to the pathogen’s transmission and persistence on farms is essential for designing effective strategies against the spread of the pathogen from farm to fork. An increasing body of evidence suggests that mobile genetic elements support the adaptation and persistence of L. monocytogenes in the food industry, as these elements contribute to the dissemination of genes encoding favorable phenotypes, such as resilience against biocides. Understanding of the role of farms as a potential reservoir of these elements is needed for managing the transmission of mobile elements across the food chain. Because L. monocytogenes coinhabits the farm ecosystem with a diversity of other bacterial species, it is important to assess the degree to which genetic elements are exchanged between Listeria and other species, as such exchanges may contribute to the rise of novel resistance phenotypes.

Agroecosystems provide a favorable habitat for L. monocytogenes, and the pathogen is especially prevalent on dairy farms (2,3). L. monocytogenes strains can inhabit dairy farms for years and be widely distributed in the farm environment, leading to the frequent contamination of milk (4,5). Raw milk and animals destined for slaughter are a major contamination source in the food industry (6)(7)(8). Knowledge of the pathogen's ecology on farms is essential for controlling the spread of L. monocytogenes from farms to the food industry.
L. monocytogenes is extremely resilient and can tolerate various stresses used in the food industry to control the pathogen (9,10). These phenotypic traits enable L. monocytogenes to survive in food processing environments for years, a phenomenon known as persistence (11)(12)(13)(14)(15). Mobile genetic elements (MGEs) are common among L. monocytogenes isolates from food processing environments (14)(15)(16) and may harbor genes mediating tolerance to heat shock (17), salt and acid stress (18,19), and biocides (20,21). These findings led us to the hypothesis that mobile genetic elements play a key role in the environmental adaptation and persistence of L. monocytogenes.
Although dairy farms are considered a reservoir of L. monocytogenes (2) and are known to harbor hypervirulent strains (22), the era of next-generation sequencing has witnessed very few efforts to illuminate the pathogen's ecology in the farm environment. How L. monocytogenes adapts to life in the farm ecosystem, and to what extent the farm environment acts as a source of mobile genetic elements for L. monocytogenes strains persisting in food processing environments, are key issues to explore. Such insights would be instrumental in developing novel strategies to reduce contamination on farms and in the raw materials delivered to the food industry.
Here, we sequenced the genomes of 250 L. monocytogenes isolates obtained from three Finnish dairy farms during 2013 to 2016 (5) to investigate the persistence and mobile genetic elements of L. monocytogenes in the farm environment. We performed a single-nucleotide polymorphism (SNP)-based phylogenomic analysis to group the isolates into persistent and nonpersistent clades and identified plasmids and chromosomal mobile elements among the 250 genomes. We found that prophages and other mobile genetic elements were, on average, more abundant among isolates in persistent clades than among those in nonpersistent clades, and that a significantly higher portion of isolates in persistent clades harbored genes against bacitracin, arsenic, and cadmium, compared to those in nonpersistent clades. Finally, we explored genomewide associations between clusters of orthologous genes and persistence. We found that defense systems against invading prophages and plasmids, including the CRISPRcas IIA system (23) and the type II restriction modification system Lmo3J (24), were negatively associated with persistence on farms. Taken together, our findings suggest that prophages and mobile genetic elements confer an ecological advantage for persistence on farms and that L. monocytogenes inhabiting the farm environment constitutes a reservoir of diverse mobile genetic elements that may spread upstream in the food chain.

RESULTS
Persisting clades of L. monocytogenes were detected on all three farms. Wholegenome sequencing and subsequent in silico subtyping of 250 Listeria monocytogenes isolates, collected from three Finnish dairy farms during 2013 to 2016 (5), yielded 25 unique multilocus sequence types (STs) ( Fig. 1a; see also Data Set S1 in the supplemental material). The most frequently detected subtype was ST20, which represented 28% of all sequenced isolates. In this study, persistent clades of L. monocytogenes were defined as monophyletic clades of isolates with pairwise distances (PWDs) of fewer than 20 SNPs (25) that were isolated from the same farm from $3 samples during $6 months. Clades that did not meet these criteria were classified as nonpersistent. In total, we identified 14 persistent clades ( Fig. 2 and Table 1). Persistent clades represented 71% of all sequenced isolates, and all persistent clades belonged to serogroup 1/2a. Clade C4 contained isolates from two different farms, suggesting that strains of L. monocytogenes can spread between farms more quickly than the rate of genomic diversification.
Pathogenicity islands associated with hypervirulence (Listeria pathogenicity island 3 [LIPI-3] and LIPI-4) were detected in 5% of the 250 isolates, none of which belonged to persistent clades. None of the 250 isolates harbored a premature stop codon within the inlA gene, which is associated with hypovirulence and is a common finding in L. monocytogenes from food processing environments (26). Indeed, the two STs most stringently associated with the food processing environment, namely, ST9 and ST121 (26), were not detected in this study.
Mobile genetic elements were on average more numerous among isolates in persistent than in nonpersistent clades of L. monocytogenes. Overall, prophages and other mobile genetic elements were significantly more numerous among isolates in persistent than among those in nonpersistent clades (P , 0.01; independent samples median test) (Fig. 1b and c). Resistance cassettes against cadmium and arsenic were detected in 20 and 15% of isolates, respectively. Mobile elements harboring resistance genes against arsenic and cadmium were significantly more prevalent among isolates in persistent clades than among those in nonpersistent clades (Fig. 3d). Surprisingly, 12% of all L. monocytogenes isolates harbored a putative bacitracin resistance cassette (27), located on the transposon Tn5801_B23. Other antimicrobial or biocide resistance genes were not detected in this study.
Dairy farm isolates of L. monocytogenes harbored plasmids that are common in the food industry and three novel plasmids. Plasmids were detected among 10% of L. monocytogenes isolates in persistent clades and 11% of isolates in nonpersistent clades. We detected three previously identified plasmids (pCFSAN010068, pLM58, and pLMR479a) and three novel plasmids, which were labeled pHC143, pHC192, and pHC195-2 ( Fig. 3a and Data Set S1). These plasmids were 55.5 to 86.7 kb in size, except for pHC192, which was only 4.6 kb. A maximum-likelihood phylogenetic analysis based on RepA grouped the five large plasmids into the plasmid groups G1, G2, and G4 (28,29), which appear to be specific to the genus Listeria (Fig. 4a). Plasmid groups G1 and G2 include well-characterized L. monocytogenes reference plasmids that are common in food processing environments (18,19,28). G4 represents a novel group of Listeria plasmids (29).
The G4 plasmid pHC143 was detected in five isolates of this study, belonging to ST6 and ST149 (see Data Set S1). These STs are hypervirulent, based on the  Recombinant sites were removed from the alignment using Gubbins 3.0. Maximum-likelihood phylogeny was inferred from concatenated SNP alignment files using PhyML 3.3. The tree was visualized using FigTree 1.4.4. Pathogenicity islands, plasmids, chromosomally located mobile elements, and prophages were identified from the assembled and annotated draft genomes. The heatmap is restricted to genomic elements that were detected in this study. Persistent clade numbers corresponding to data in Table 1  presence of pathogenicity islands LIPI-3 (ST6) and LIPI-4 (ST149) (Fig. 2). Visualization of assembly graphs indicated that pHC143 was successfully assembled into a single 55.8-kb contig in all five isolates. pHC143 contained no biocide or heavy metal resistance genes. However, we identified three variants of pHC143 among short-read sequence assemblies deposited in GenBank, all of which contain resistance genes against biocides ( Fig. 4b; see also Fig. S1 in the supplemental material). The first variant contains a benzalkonium chloride resistance cassette (bcrABC) and a mercuric resistance (mer) operon. The second variant contains a multidrug exporter putatively conferring resistance against quaternary ammonium compounds (qacC [qacH]; GenBank accession number WP_000121134.1). The third variant contains the qacC/qacH gene and a Tn554 family transposon carrying an arsenic resistance operon (arsABCD). This Tn554 family transposon was identified previously in the chromosomes of L. monocytogenes (30). All G4 plasmids contained a predicted fimbrial adhesin (GenBank accession number WP_061691480.1), suggestive of a role associated with attachment and host colonization (31).
Assembly graphs of the small plasmid pHC192 suggested that the plasmid was closed successfully into a single 4.6-kb contig. pHC192 did not contain replication proteins related to the RepA of Listeria plasmid groups G1 to G4, so the phylogeny of this plasmid was analyzed using RepB (Fig. 5a). Phylogenetically, pHC192 clustered closely with plasmids from Lactobacillus. Indeed, RepB of pHC192 (GenBank accession number WP_035147907.1) was also detected in Lactobacillus and Brochothrix (100% amino acid sequence identity), suggestive of a broad host range for this plasmid. The closest relative of pHC192 in Listeria was the plasmid of L. monocytogenes strain CFIAFB20130002, which possesses the lincosamide resistance gene lnuA (GenBank accession number WP _001829870.1). Notably, RepB of pHC192 bore no similarity to the replication proteins of the small Listeria plasmids pIP823 (GenBank accession number WP_172694646.1) and pDB2011 (accession number WP_020277964.1) and shared only 45% amino acid identity with the RepB of pLMST6 (accession number WP_061092472.1). Like pHC192, pLMST6 appears to also have a broad host range, as 100% identical homologues of pLMST6 RepB (accession number WP_061092472.1) were detected in Listeria, Salmonella, and Enterococcus. These findings suggest that several phylogenetically unrelated small plasmids have been acquired by Listeria through distinct transfer events across host species. The plasmid pHC192 contains a putative tauE (safE) family sulfite exporter gene (GenBank accession number WP_016896343.1) (Fig. 5b) that is not typically present in Listeria plasmids (19,29). The sequencing depth of coverage for pHC192 was approximately five times that of the chromosome, suggesting that pHC192 is a high-copynumber plasmid. This plasmid became increasingly prevalent among persistent clade C7 isolates during the sampling period and was detected in all isolates at the end of the study (Fig. 5c). An additional plasmid, pHC195-2, was detected in several isolates of clade C7 in the latter part of the sampling period. The pHC195-2 plasmid belonged to the phylogenetic group G2 (Fig. 4a) and closely resembled the reference plasmid   (29). Tip labels correspond to plasmid names and host genera; plasmids from this study are labeled in blue. (b) G4 plasmids of the L. monocytogenes strains HC193 (this study), HC374 (this study), and FDA550584-30 (BioSample accession number SAMN02923676) aligned with .95% identity across the entire length of pHC143 from this study; plasmids of the L. monocytogenes strains 967535 (BioSample accession number SAMN15680309) and YA00079283 (accession number SAMN08970420) aligned with .95% identity to most of pHC143. Red arrows indicate the insertion sites of the biocide resistance loci qacC, present in p967535 and pYA00079283, and bcrABC, present in pFDA550584-30. The alignment was generated using BRIG 0.95. For pHC143, plasmid length in base pairs (bp) is given. pLMR479a (see Fig. S2 in the supplemental material). The acquisition of these plasmids during the course of persistence suggests that they play a role in the adaption of the pathogen to the farm ecosystem.
The IS3-like transposon was significantly more prevalent among isolates in persistent than in nonpersistent clades (Fig. 3b). The IS3-like transposon consists of two insertion sequences in lineage I (IS3-1 and IS3-2) and a single insertion sequence in lineage II (IS3-1) (Fig. 2). These elements harbor multiple surface-associated lipoproteins, which may facilitate attachment and invasion (30). The suggested role of the IS3-like transposon in L. monocytogenes virulence remains to be determined.
LGI-2 carries cadmium and arsenic resistance cassettes and two multidrug transporters (see Fig. S3 in the supplemental material). Identical (100% nucleotide identity) LGI-2 elements were present among all ST14 and ST145 isolates in this study (Fig. 2). Moreover, a BLASTn search identified identical LGI-2 elements in 11 L. monocytogenes and two Enterococcus faecalis complete genomes, suggestive of recent transfer between these species.
Tn5801_B23 was detected in a subset of ST20 isolates, including the persistent clades C9 to C12 (Fig. 2 and Data Set S1). The Tn5801_B23 detected in this study shared 97% nucleotide identity with the Tn5801_B23 of Enterococcus faecalis strain JH2-2 (see Fig. S4 in the supplemental material). Tn5801_B23 contains putative resistance genes against the antimicrobial bacitracin (bcrABD) and a two-component system (baeSR) potentially involved in the regulation of the bcrABD operon (33). Unlike Tn5801_B23, other Tn5801like elements mediate tetracycline resistance in Enterococcus, Listeria, and several other Firmicutes species (33). In L. monocytogenes ST20, Tn5801_B23 was inserted downstream of guaA (lmo1096), which is also the insertion site of the related element ICELm1 of L. monocytogenes strain EGD-e, harboring cadmium resistance genes (30).
The putative integrative and mobilizable element (IME) Tn7101 was detected in the ST155 singleton isolate HC258, where it was inserted between homologues of lmo2596 and lmo2597 (see Fig. S3). Tn7101 contains resistance genes against cadmium (cadA and cadC) and an arsenate reductase (arsC). Through a BLASTp search, we identified a variant of Tn7101 containing a seven gene arsenic resistance cassette. This variant, labeled Tn7102, was detected in several L. monocytogenes and Enterococcus genomes deposited in GenBank (see Fig. S2). The Tn7101 and Tn7102 of Listeria and Enterococcus were identical (100% nucleotide identity), suggestive of recent promiscuity between the two genera. Arsenic resistance genes in Tn7102 were distantly related ($67% nucleotide identity) to the arsenic resistance cassette of LGI-2 (see Fig. S3).
The putative IME Tn7103 was detected in the ST119 singleton isolate HC183, where it was inserted between lmo0810 and lmo0811. This transposon contained putative virulence genes encoding an InlJ-like internalin and a bacterial immunoglobulin (Big)-like protein (see Fig. S5 in the supplemental material). A BLAST search confirmed the presence of Tn7103 in other L. monocytogenes strains, including N12-2532 (BioSample accession number SAMN09947958), but we did not identify this element in other species.
The putative ICE Tn7104 was detected in the ST391 singleton isolate HC187 and was inserted between lmo1786 and lmo1787. This transposon contained a putative type I restriction modification system (see Fig. S6 in the supplemental material). Tn7104 was identified in several other L. monocytogenes strains deposited in GenBank, including the L. monocytogenes ST391 strain SHL013 (BioSample accession number SAMN03265960), but we did not identify this element in other species.
A novel prophage harboring cadmium resistance genes was identified in a persistent clade of L. monocytogenes. All 250 dairy farm isolates from this study contained the L. monocytogenes monocin (39) and 0 to 3 additional prophages, which were detected at eight insertion sites (Fig. 3c). Prophages inserted into tRNA-Arg(tct) were significantly more prevalent among isolates in persistent clades, and prophages inserted into tRNA-Lys(ctt) were significantly more prevalent among nonpersistent clades (P , 0.05; Fisher's exact test).
OPTSIL taxonomic clustering assigned prophages from this study into six genera. Prophages inserted into comK and tRNA genes were assigned to genera of Siphoviridae that are known to only infect Listeria. Surprisingly, in the isolate HC189, a 67-kb Myovirus was inserted into comK, a site usually occupied by Siphoviridae (40).
Prophages inserted between the rlmCD (lmo1703) and fosX (lmo1702) genes were not related to any of the Listeria-specific phage genera, but instead represented a separate genus that infects several Firmicutes species (Fig. 6). Many of the phages in this genus harbor antimicrobial and heavy metal resistance cassettes (see Fig. S7 in the supplemental material). In this study, phages inserted between rlmCD and fosX were detected among all isolates of persistent clade C8 and among three singleton isolates (Fig. 2). Among isolates of persistent clade C8, prophages inserted between rlmCD and fosX all harbored a cadmium resistance cassette (see Fig. S7). In contrast, in the singleton isolates, prophages inserted between rlmCD and fosX harbored no cadmium or antimicrobial resistance genes. Within Listeria genomes deposited in GenBank, we identified prophages inserted FIG 6 Prophages inserted between rlmCD and fosX belonged to the genus Siphovirus and have a broad host species range and a tendency to harbor antimicrobial or heavy metal resistance genes. Minimum evolutionary tree and taxonomic clustering of six Listeria-specific phages (genera 2 to 6), prophages from the strains HC134 and HC258 from this study that were inserted between rlmCD and fosX (genus 1, blue), and related prophages from Listeria and other Firmicutes species obtained from GenBank (genus 1, black). Phylogenetic analyses and clustering were generated with the VICTOR online tool (https://victor.dsmz.de), using model D 6  between rlmCD and fosX that carried resistance genes against cadmium (cadA), macrolides (mefA, msrD), tetracycline (tetM), and streptogramin (vatA).
Systems that protect against invading DNA were negatively associated with the persistence of L. monocytogenes on dairy farms. A genome-wide association study was conducted to assess which genes were associated with persistent versus nonpersistent clades. Because no persistent clades belonged to lineage I, the analysis was restricted to lineage II. Notably, a gene involved in biofilm formation (bapL), which has been implicated in the adaptation of L. monocytogenes to the food processing environment (22), was significantly associated with persistence on dairy farms (see Table S2 in the supplemental material). In contrast, genes associated with the CRISPR-cas type IIA system and the type II restriction-modification system LmoJ3 (24) were negatively associated with nonpersistence (see Table S2). CRISPR-cas systems and restriction modification systems may act in synchrony to protect the host against invading prophages and other mobile elements (41). Additionally, a putative recombination and DNA strand exchange inhibitor protein (GenBank accession number WP_031664941.1) was associated with nonpersistence. These findings agree with the lower prevalence of mobile genetic elements and prophages among isolates in nonpersistent than those in persistent clades and suggest that systems involved with inhibiting invading DNA are detrimental for the persistence of L. monocytogenes in the dairy farm environment.
Curiously, the ESX-1-like type VII secretion system (T7SS) contained both genes associated with persistence and genes associated with nonpersistence. The T7SS of L. monocytogenes has a potential role in bacterial antagonism (42) and is located in L. monocytogenes hypervariable hot spot 1 (30). Overall, many of the genes associated with persistence or nonpersistence belonged to L. monocytogenes hypervariable hot spots or prophages, suggesting that the role these components play in Listeria niche adaptation deems further study.

DISCUSSION
Whole-genome sequencing and subsequent analyses of 250 L. monocytogenes isolates from dairy farms illustrated that dairy farm isolates are hosts to a diversity of mobile genetic elements that carry, or have the potential to carry, resistance genes against antimicrobials, biocides, and heavy metals. Many of the mobile elements we identified carried genes encoding phenotypes that promote the survival of L. monocytogenes on farms, such as antimicrobial resistance genes or virulence factors. Moreover, genes responsible for the conjugation of mobile elements may have a dual role in promoting biofilm formation and invasion of the mammalian host (31,43), further enhancing the survival of L. monocytogenes on farms. We found that prophages and other mobile genetic elements were significantly more numerous among isolates belonging to persistent than nonpersistent clades. Moreover, systems that provide immunity against invading mobile genetic elements (23,24,41), namely, the CRISPR-cas IIA system, the type II restriction modification system LmoJ3, and a putative recombination and DNA strand exchange inhibitor protein, were associated with nonpersistence. Together, these findings suggest that mobile elements may support the persistence of L. monocytogenes inhabiting farms.
Most of the mobile genetic elements we uncovered appeared in a very limited number of L. monocytogenes STs. The narrow distribution and wide diversity of the mobile genetic elements we identified likely explain why very few mobilome-associated genes were significantly associated with persistence in this study. As an exception, the IS3-like element of L. monocytogenes (30) and prophages at certain insertion sites were widely distributed across STs and harbored genes that were significantly associated with either persistence or nonpersistence.
We identified a surprising diversity of mobile genetic elements encoding heavy metal resistance genes among the dairy farm isolates. Moreover, acquired heavy metal resistance genes were more common among isolates in persistent than nonpersistent clades. Heavy metal resistance is also more common among persistent than nonpersistent L. monocytogenes subtypes from foods and food processing environments (14,16). Whether heavy metal resistance genes contribute directly to persistence or merely cooccur with other determinants that promote environmental survival remains unclear (44). Nevertheless, heavy metal resistance genes may represent useful markers to aid the detection of L. monocytogenes strains with enhanced resilience against environmental stressors.
In the present study, we found a novel plasmid (pHC143; plasmid group G4) that infected hypervirulent subtypes of L. monocytogenes. Although pHC143 was devoid of biocide and heavy metal resistance genes, such genes are common on other G4 plasmids infecting hypervirulent ST1 and ST6 strains (29). Indeed, we noted a G4 plasmid harboring the biocide resistance gene qacC and the arsenic resistance cassette arsABCD in the ST6 outbreak isolate YA00079283, associated with the largest listeriosis outbreak known to date (45). It is plausible that harborage of biocide and heavy metal resistance genes in G4 plasmids facilitates the adaptation of hypervirulent strains to food processing environments.
We identified four transposons in Listeria, namely, LGI-2, Tn5801_B23, Tn7101, and Tn7102, that closely resembled transposons in Enterococcus, suggestive of recent transfer between the two genera. The cooccurrence of genomic elements in Enterococcus and Listeria was unsurprising, as both genera are highly prevalent in animal feces and farms (5,8,46). Transfer of conjugative elements has been demonstrated both from Enterococcus to Listeria and vice versa (47,48) indicating that both genera are potential donors. The extent to which enterococci and other Firmicutes contribute to the horizontal spread of mobile elements and their associated antimicrobial, biocide and heavy metal resistance determinants in Listeria has implications for food safety and should be explored through further study.
Bacitracin resistance genes, mediated by Tn5801_B23, were common among L. monocytogenes isolates from all three farms investigated. Moreover, Tn5801_B23 was significantly more prevalent among isolates in persistent clades than among those in nonpersistent clades. The widespread use of bacitracin as a growth promoter in animal feeds has facilitated the expansion of bacitracin resistance in Enterococcus (49,50), and probably also in L. monocytogenes, as animal feeds are frequently contaminated by Listeria (5, 8). Nevertheless, the frequent detection of Tn5801_B23 in this study remains curious, as feed supplementation with bacitracin subsided in Finland in the 1990s (49).
We found that prophages were more prevalent among isolates in persistent clades than among those in nonpersistent clades. Whether prophages contribute to the persistence of L. monocytogenes is an intriguing possibility. There is increasing evidence that prophages can mediate beneficial phenotypes for their host. Phages mediate resistance or virulence properties in numerous bacterial species (51), and in Listeria, siphoviruses inserted into comK were found to regulate the gene in a symbiotic manner (40). Here, we discovered phage-mediated carriage of cadmium resistance and various antimicrobials in Listeria, suggesting that prophages contribute to the spread of phenotypes supporting persistence. Moreover, we noted that these phages belonged to a genus of Siphovirus with an apparently broad host species range that were introduced to Listeria through several distinct transfer evets. Host species jumps have the potential accelerate the transfer of novel resistance determinants between Listeria and other Firmicutes.
It is worth noting that not all persistent clades harbored mobile elements, suggesting that other factors also contribute to the survival of L. monocytogenes on dairy farms. We found that genes putatively involved in biofilm formation (bapL) and interbacterial competition (T7SS), which are not located in mobile elements, were significantly associated with persistence. In addition, the predominance of persistent L. monocytogenes strains in the dairy farm environment is associated with inadequacies in production hygiene (5). Therefore, the persistence of L. monocytogenes in the dairy farm environment is likely the result of a multifactorial combination of bacterial and environmental factors.
In conclusion, our study indicates that L. monocytogenes strains inhabiting the dairy farm environment are receptive to a diversity of prophages and mobile genetic elements. We suggest that mobile elements enable L. monocytogenes to adapt to the stresses encountered in the farm ecosystem and in general improve the fitness of the pathogen on farms, thereby supporting persistence. Given the abundance of L. monocytogenes on farms (2,3,5) and the apparent exchange of mobile genetic elements between Listeria and other Firmicutes species, L. monocytogenes occurring in agroecosystems should be viewed as a potential reservoir of mobile genetic elements. Importantly, many of these elements have the potential to carry and spread antimicrobial, biocide, and heavy metal resistance genes. The spread of mobile genetic elements and resistance determinants from primary production to Listeria in food processing environments has important food safety implications and should be explored further. The present study represents a step forward in this effort and in our understanding of listerial ecology in the agroecosystem.

MATERIALS AND METHODS
Whole-genome sequencing. In total, 250 L. monocytogenes isolates obtained from three Finnish dairy cattle farms during 2013 to 2016 (5) were selected for whole-genome sequencing in the present study (see Data Set S1 in the supplemental material). The isolates were obtained from samples of bulk tank milk (31 isolates), used milk filters (46 isolates), feed (14 isolates), cow feces (21 isolates), and bedding materials (9 isolates) and from surface swab samples of floors (54 isolates), feed and water troughs (54 isolates), udders and udder cloths (10 isolates), milking systems and bulk tanks (8 isolates), stall mats (2 isolates), and strip cups (1 isolate). DNA was extracted from overnight cultures using the guanidium thiocyanate extraction method (52). DNA samples were standardized to a concentration of 10 ng/ml using the double-stranded DNA (dsDNA) broad-range (BR) assay kit (Thermo Fisher Scientific, Waltham, MA) using a Qubit fluorometer (Thermo Fisher Scientific). Genomic libraries were constructed from the DNA samples using the Nextera XT DNA sample preparation kit (Illumina, San Diego, CA), and pairedend sequencing (2 Â 250 bp) was performed using the Illumina HiSeq 2500 platform.
Genome assembly, pangenome construction, and subtyping. Following the removal of adapter sequences and low-quality reads using Trimmomatic 0.36 (53), draft genomes were assembled using SPAdes 3.9 with k-mer values of 55, 77, 99, 113, and 127 (54). Assembly quality was assessed using QUAST 4.0 (55), and taxonomic assignment was performed using Kraken (56). The assemblies were annotated using Prokka 1.12 (57). The pangenome of the sequenced isolates was constructed using Roary 3.8.0 (58) with the protein identity cutoff value set at 90%. Multilocus sequence types (ST), corresponding to the schema developed by Ragon et al. (59), and core genome sequence types (CT), corresponding to the schema developed by Moura et al. (60), were determined in silico from the assembled genomes using the BIGSdb-Lm database. The BIGSdb-Lm database was also used to identify pathogenicity islands associated with hypervirulence (LIPI-3 and LIPI-4) and genes associated with antimicrobial and biocide resistance among the assembled genomes. Genome assemblies were deposited in GenBank under BioProject accession number PRJNA704814 (see Data Set S1).
Maximum-likelihood phylogenomic analysis. Phylogenomic reconstruction of the 250 L. monocytogenes isolates was performed using the Lyve-SET 1.1.4f pipeline (61), using the L. monocytogenes EGDe genome (GenBank accession number NC_003210.1) as a reference. The Lyve-SET pipeline was run using Listeria monocytogenes presets (61), with the additional options "-mask-phages," "-mask-cliffs," and "-read_cleaner CGP." In brief, the pipeline generated genome alignments by mapping quality-filtered reads to a reference genome. To improve the accuracy of phylogenomic inference, putative prophage genes were removed from the reference genome prior to mapping. Mapping was followed by the detection of high-quality SNPs, defined as having $10Â depth of coverage and $75% consensus among reads. Recombinant sites within the genome alignments generated by Lyve-SET were identified and removed using Gubbins 3.0 (62). PhyML 3.3 (63) was used to infer maximum-likelihood phylogeny of each ST using a general time-reversible model (GTR) with 100 bootstrap replicates.
In addition, the phylogeny of each ST harboring putative persistent clades was inferred independently. Persistent clades of L. monocytogenes were defined as monophyletic clades of isolates with PWDs of ,20 SNPs (25) that were isolated from the same farm from $3 samples during $6 months. For each ST, a draft assembly from the present study with the best quality statistics, i.e., the highest N 50 value and lowest number of contigs (see Data Set S1), was used as a reference genome. The phylogenomic analyses were executed as described above using the Lyve-SET pipeline, Gubbins, and PhyML.
Detection and analysis of plasmids. Plasmids were identified by aligning the whole-genome assemblies against Listeria plasmids deposited in GenBank with the aid of BLASTn (http://www.ncbi.nlm .nih.gov/blast). Alignments were inspected manually. Additionally, whole-genome assembly graphs generated by SPAdes were visualized using Bandage 0.8.1 (64), and extrachromosomal elements were inspected manually. Maximum-likelihood phylogeny of the plasmids, based on the amino acid sequence alignments of the repA gene, was generated with MEGA7 (65), using the Jones-Taylor-Thornton substitution model with 100 bootstraps. Alignments of the amino acid sequences of the repB gene were used to compare plasmids in which repA was absent. Plasmid alignments were generated and visualized using BRIG 0.95 (66) and EasyFig 1.2 (67).
Detection and analysis of prophages. Prophages inserted into the L. monocytogenes genomes were identified using PHASTER (68), and the insertion sites were inspected manually. Phylogeny and taxonomic clustering of prophages classified by the PHASTER algorithm as "intact" were inferred using VICTOR (69). Nineteen additional Listeria phage genomes and one streptococcal phage genome obtained from GenBank were included in the analyses for reference (see Table S1). In brief, VICTOR applies the genome BLAST distance phylogeny (GBDP) method (70) to obtain pairwise distances, from which balanced minimum evolution trees are inferred. VICTOR utilizes OPTSIL (71) to obtain taxonomic clustering. Duplicate phage genomes are removed from the analysis. Trees generated by VICTOR were visualized using FigTree 1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/). BLAST was used to identify phages inserted between rlmCD and fosX in the genomes of Listeria and other bacterial species deposited in GenBank, and hits were inspected manually. Phylogeny and taxonomic clustering of prophages inserted between rlmCD and fosX were inferred using VICTOR.
Identification of genes associated with predominance. Scoary 1.6.16 (72) was used to identify genes that are significantly associated with occurrence in persistent versus nonpersistent clades. Scoary was executed using default options, using the "gene_presence_absence.csv" file generated by Roary as the input. Associations with a Bonferroni-corrected P value of 0.05 were considered significant. As all persistent clades belonged to lineage II, the analysis was limited to the 233 lineage II isolates of this study to reduce noise arising from population structure bias.
Data availability. The 250 L. monocytogenes isolates sequenced and analyzed in this study have been deposited in the NCBI BioSample database under accession numbers SAMN18056206 to SAMN18056455 and in GenBank under BioProject accession number PRJNA704814 and are described further in Data Set S1 in the supplemental material.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. DATA SET S1, XLSX file, 0.1 MB.