Genomic analysis of avian-pathogenic Escherichia coli (APEC) isolated from diseased chicken

Background Avian pathogenic Escherichia coli (APEC) can cause various extraintestinal infections in chicken, resulting in massive economic losses in poultry industry. Apart from that, some avian E. coli strains may have zoonotic potential, making poultry a possible source of infection for humans. Due to its extreme genetic diversity, this pathotype remains poorly defined. This study aimed to investigate the diversity of colibacillosis-associated E. coli isolates from Central European countries with a focus on the Czech Republic. Results Out of 95 preliminarily characterized clinical isolates 32 isolates were selected for whole-genome sequencing. A multiresistant phenotype was detected in a majority of them and the predominant resistance to lactams and quinolones was widely associated with TEM-type beta-lactamase genes and chromosomal gyrA mutations, respectively. The phylogenetic analysis confirmed a great diversity of isolates, that were derived from nearly all phylogenetic groups, with predominace of B2, B1 and C phylogroups. Clusters of closely related isolates within ST23 (phylogroup C) and ST429 (phylogroup B2) indicated a long-term local spread of these clones. Besides, the ST429 cluster carried bla CMY-2, -59 genes for AmpC beta-lactamase and isolates of both clusters were generally well-equipped with virulence-associated genes, with considerable differences in distribution of certain virulence-associated genes between phylogenetically distant lineages. Other important and potentially zoonotic APEC STs were detected, incl. ST117, ST354 and ST95, showing several molecular features typical for human ExPEC. Conclusions The results support the concept of local spread of virulent APEC clones, as well as of zoonotic potential of specific poultry-associated lineages, and highlight the need to investigate the possible source of these pathogenic strains.


Background
Avian colibacillosis is a complex of several localized or systemic syndromes, affecting poultry of all age and production categories. It comprises yolk sac infection and omphalitis, leading to increased mortality rates in newly hatched chicks, cellulitis in broilers or reproductive tract infections in laying hens. Other forms of manifestation include swollen head syndrome (SHS), respiratory infections and septicemia which frequently result in death or chronic forms of infection. Avian colibacillosis thus represents a great economic burden for the poultry industry (1). Despite its importance, pathogenesis of these infections is rather intriguing and not well understood. For a long time APEC strains were considered merely opportunistic pathogens, predominantly, but not exclusively associated with O1, O2, O8, O78 and several other serogroups (2). It has nevertheless been proved that disease-associated E. coli strains encode multiple putative virulence genes and significantly differ from commensals, particularly in the presence of ColV plasmid-associated genes, possible markers of poultry-adapted pathogenic strains (3)(4)(5).
The ability to cause colibacillosis in chicken defines the APEC (avian-pathogenic E. coli) pathotype. However, not every strain isolated from diseased chicken carries typical virulence-associated genes, underlining an opportunistic character of some types of E. coli infections (6). On the other hand, APEC can be found also in the gut of healthy chicken (7,8). It has been suggested by Maturana et al. (9) that APEC population composes of distinct subpathotypes associated with different syndromes, similar to the human extraintestinal pathogenic E. coli (ExPEC). Interestingly, the authors showed that SHS and omphalitis isolates formed two distinct groups differing in virulence, suggesting primary and opportunistic character of those infections, respectively. Similarly, chronic salpingitisperitonitis syndrome resulting from an ascending infection and an acute peritonitis 4 without salpingitis, probably originating from respiratory infection or gut translocation after a stress insult, can be distinguished in layers (10)(11)(12).
There is a close genetic relationship between APEC and human ExPEC. Zoonotic potential of poultry strains has been implicated. ExPEC are the main cause of urinary-tract infections (as so called uropathogenic E. coli, UPEC) in humans and meningitis in neonates (neonatal-meningitis E. coli, NMEC), and are also associated with bacteremia, sepsis, cellulitis and other sometimes fatal infections (13). Similar to APEC, these strains are characterized by the presence of various virulence-associated genes, participating in adhesion and colonization of different tissues, invasion of internal organs, iron acquisition and avoiding host's immune responses. ExPEC are typically associated with the phylogenetic groups B2 and D, in contrast to commensal and intestinal pathogenic strains derived from groups A and B1 (14) and to APEC, which are highly variable in distribution to various phylogenetic groups (15). Although there is no specific set of genes to define the subpathotypes (16,17), APEC, UPEC and NMEC generally form genetically distinct groups. There is, however, a substantial overlap especially within the B2 phylogenetic group, which comprises strains isolated from both humans and chickens, showing high virulence in both chicken and mammalian models with low or no host specifity (18,19,17).
Moreover, an isolate showing high virulence in rat meningitis model has been found in faeces of a healthy chicken (5), another finding implying a potential importance of poultry and poultry products as a source of human pathogens.
Recently, several highly virulent and resistant ExPEC lineages with worldwide distribution have emerged (e.g. ST131, ST95 etc.) (20). Whereas some of them are associated exclusively with human infections, others are frequently isolated from diseased poultry or poultry products (21)(22)(23)(24)(25)(26). It is however difficult to assess the real importance of poultry as a source of human infections. Mechanisms of transmission of pathogenic clones through 5 the production chain to humans are very complex and not quite elucidated, as well as the relationships between genetic "armory" of virulence and resistance-associated genes and pathogenesis of the disease. Whole-genome sequencing (WGS) represents a revolutionary tool to study these mechanisms in their complexity (27). Moreover, an immense variability of APEC pathotype and differences in geographic distribution of specific clones underlines the importance of mapping the local situation. While several papers have reported occurrence of highly pathogenic APEC clones in different counries, the information for the Central Europe have been sporadic or is lacking (28).

Samples cellection and preliminary characterization
A total of 95 isolates were subjected to preliminary characterization including serogrouping, antimicrobial resistance (AMR) testing and PCR detection of virulence and antibiotic resistance genes. The disc diffusion test showed that 69.5% were resistant to three or more groups of antimicrobials, which we considered as a criterion of multiresistance. Resistance to ampicillin was recorded in 78 isolates (82.0%), followed by resistance to nalidixic acid (62 isolates; 65.3%), sulphonamides (45; 47.4%) and sulphonamides-trimethoprim (28; 29.5%). Nineteen isolates (20.0%) showed reduced susceptibility to ciprofloxacin (additional file 1- Figure 1). Using four antisera (O1, O8, O18 and O78), 49 isolates (52%) were typeable, with predominant serogroups O1 (30; 32%) and O8 (13; 14%). Only four, respectively two isolates reacted positively against antisera O78 and O18. isolates failed to be typed by WGS (the results are summarized in Table 2, supplementary material B). Except for serogroup O8 (7 isolates), the remaining serogroups were only represented by one or two isolates. The predominance of O8 serogroup appeared as a selection bias since only Czech isolates were selected for sequencing. Of the O8 serogroup, six isolates belonged to the O8:H9 serotype, most of them to ST23 type. In two cases (O1 and O78) the isolates reacted positively in the agglutination test, but gene identity of in silico identified genes was below threshold.
The core genome consisted of 2763 genes (55.28 kbp). The phylogenetic tree based on the core SNPs analysis basically corresponded to the structure of E. coli phylogeny. Groups F, D and clade I were represented by only a few isolates and did not form any distinct clusters except the minor subcluster of the two group D isolates; two ST117 isolates from the F phylogroup were unrelated to other isolate of F phylogroup (ST354) and formed their own distinct clade. In the B2 cluster two subclusters (B2a, B2b) were found; B2b subcluster was formed by four closely related ST429 isolates and one ST4110. Another cluster included isolates from phylogroups A, C and B1. Interestingly, all isolates of the C group belonged to ST23, O8:H9 serotype (with one exception of O78:H9 serotype). (1/32; 3.0%) and bla TEM-30 (1/32; 3,0%). In addition, all isolates showed the presence of genes encoding components of various multidrug efflux pumps, participating in resistance to aminoglycosides, macrolides and fluoroquinolones. Except for qnrS1, which is associated with partial resistance to fluoroquinolones, no other PMQR gene was detected.

Identification of resistance genes
Reduced susceptibility to quinolones in most isolates appeared to be due to chromosomal mutations, especially in gyrA gene (21; 65.6%), to lesser extent also in parC (5; 15.6%) and parE (1; 3%). In five ST23 isolates (15.6%), a mutation in ampC promoter was 8 detected. (For overview of resistance genes, please see the Table 1, additional file 1.)

Identification of virulence genes
The genomic analysis confirmed a great diversity of selected isolates (see figure 2 and supplementary material, file 3). Overall, factors associated with adhesion and invasion, as well as siderophores were found in most isolates; more than 90 % of isolates encoded F1 fimbriae, curli, E. coli common pilus and enterobactin. All but one isolate carried ibeB gene, while ibeA was present mostly in B2 and F phylogenetic groups, but not in isolates from other groups. A siderophore system salmochelin (81%) and serum-resistance associated proteins, Iss (87.5%) and TraT (78%) were present in most isolates with generally equal distribution in all phylogenetic groups. Full SitABCD iron transport system was detected in 78% isolates, outer membrane protease (OmpT) and colicin V synthesis protein (CvaC) in 68.8% and 59% isolates, respectively.

Discussion
The aim of current study was to evaluate diversity of colibacillosis-associated isolates from Central Europe. Indeed, analysis showed an immense phenotypic and genotypic variability, the isolates differing greatly in their antimicrobial-resistance phenotype, virulence genes profile and plasmid content, together having little in common. As generally acknowledged, there is no specific combination of virulence genes that would accurately define the APEC pathotype (16). The most prevalent APEC genes are also frequently present in commensal strains. There is an abundance of adhesins and irontransporting systems, which may be considered essential prerequsites of extraintestinal pathogenicity in all types of avian and mammalian disease, but also fitness factors enabling asymptomatic colonization of healthy hosts and effective transmission. Presence of Col-V-associated genes such as iroN, iss, iutA, ompT etc. is characteristic for most APEC, more than UPEC and NMEC (15), nevertheless, their exact role in pathogenesis remains unclear or controversial (30,31). Col-V-like plasmids are, however, acknowledged as markers of poultry-adapted pathogenic strains (5,22).
As expected, the phylogenetic analysis also revealed a substantial diversity of isolates, that originated from all phylogenetic groups with the exception of group E, the most prevalent was B2 phylogroup, which is, along with D, considered typical group for human ExPEC (14). However, the second most prevalent phylogroup was B1 (7 isolates), a group commonly associated with intestinal pathogenic or commensal fecal strains (32).
Interestingly, there were notable differences in virulence trait distribution among phylogenetic groups, although the isolates had been collected from the same type of infection. The idea of pathogenic strains with quite different combination of virulence genes with alternative function causing the same clinical disease has been proposed by Mokady et al. (33) and points out the importance of horizontal gene transfer enabling rapid adaptation to new niches by expression of certain genes in a different genetic background (34). Notably, it was the presence of typical Col-V plasmid-associated genes such as ompT, iss, cvaC, iro and sit (but suprisingly not iut, iuc for aerobactin) that were equally distributed among isolates from all phylogenetic groups.
Despite the overall diversity, the phylogenetic analysis revealed two clusters (ST429, group B2, and ST23, groups C), both containing four similar isolates that were obtained from different farms in Northern Moravia. Two ST23 isolates identical according to the core genome analysis were collected at the same day on two different farms, indicating a possible clonal spread in the locality. Colibacillosis outbreaks caused by a specific pathogenic clone have been repeatedly reported (e.g. 12,35). On the other hand, a closely related isolate (25 SNPs difference) had been collected on an unrelated farm approximately half a year before. Similar situation was observed in the ST429 cluster-the most similar isolates were from the same date and were separated from the other isolates of this cluster (with 26-61 SNPs difference) by a span of several months. One may imagine these isolates could have a common origin, however, the question, whether these clones may become established somehow in the production chain and circulate between flocks or farms for a long time period or whether a repeated introduction occurs from a specific source, remains unanswered. An evidence for "pseudo-vertical" spread through the production pyramid has been proposed recently (36). The problem of possible reservoir of pathogenic strains for Northern Moravian farms should be addressed more closely in the future.
Both ST429 and ST23 are considered as predominant APEC lineages that are frequently isolated from poultry with clinically manifested disease (35,37), but also poultry products (26). Although representing quite unrelated APEC clades, they both appear to be poultryspecific, with little pathogenic potential for humans (7,17). In fact, an APEC strain χ7122 (ST23) has been shown to be phylogenetically closer to human ETEC (without any enterotoxin production) than to ExPEC (38). Therefore, in our collection, one may consider the two clusters, ST429 and ST23, representatives of phylogenetically distant lineages presumably associated with the same disease, again underlining the importance of accessory genome in virulence potential of APEC. The ST429 isolates had a slightly higher average number of virulence-asociated genes than ST23 isolates (172 vs. 154) including genes encoding capsule production (kpsM, T, D, neuC),, invasins (ibeA, ompA) and ironbinding systems (aerobactin, yersiniabactin, chu) that the ST23 cluster (not all ST23 isolates) lacked. In contrast, ST23 isolates were characterized by presence of Stg fimbriae and ETT2-related genes. This transport system, even in degenerate state, has been reported to enhance virulence in APEC (39). Both sequence types coded for curli, F1 fimbriae, salmochelin, OmpT, TraT, Iss, however, only Iro, OmpT nad Iss have been reported to occur in significantly higher prevalence in APEC than avian-faecal E. coli (AFEC) (4). Nevertheless it probably supports the idea of feasibility and usefulness of PCR typing targeting such potential markers of APEC derived from distant phylogenetic groups (e.g. 3).
Two isolates were assigned to ST117 (phylogenetic group F). Recent studies indicate that this sequence type comprises an important APEC lineages that are repeatedly reported from colibacillosis outbreaks in different countries (25,37,(40)(41)(42), but are also highlighted as potential zoonotic pathogens for containing ExPEC-related virulence genes and being isolated from both retail poultry meat and human clinical urinary tract infections (43). The remaining phylogroup F isolate was ST354, another potentially zoonotic ST, reported particularly from human and animal healthcare facilities and characterized by common resistance to antimicrobials including fluoroquinolones (44,45). This isolate carried bla CMY-2,-59 and mulitple adhesin genes including K99/F5 fimbriae, which were not found anywhere else. Both ST117 and ST354 were highly prevalent among ESBL/AmpC positive chicken isolates and it has been proposed that these lineages exhibit effective host colonization and persistence in the environment (41,45).
ST95 is probably the most important pandemic ExPEC lineage that is frequently isolated from chickens (23,25,46). In fact, it may represent, along with closely related ST140, that part of B2 phylogroup where human ExPEC and APEC form a single "subpathotype" of genetically indistingushable strains (17)(18)(19). In humans, ST95 was associated with bloodstream infections, UTIs and meningitis, often characterized by serogroups O1, O2, O45, flagellar antigen H7 and K1 capsule (typical feature of NMEC) and, in contrast to other pandemic lineages, relatively low tendency to acquire antimicrobial resistence (20,47). Indeed, not every ST95 seems to be zoonotic, as was shown with APEC O1 in a murine infectious model (48). On the other hand, our ST95 isolate fulfilled the molecular criteria for UPEC as defined by Johnson et al. (49).
Antimicrobial-resistance profile ranged from full susceptibility to all antimicrobials tested to multidrug resistance, with dominating resistance to β-lactams (ampicillin) and first generation quinolones (nalidixic acid). Resistance to β-lactams was associated largely with TEM-type β-lactamase production. No selection procedure to obtain ESBL/AmpC producing isolates had been used and we did not detect any bla CTX-M gene, while four isolates carried bla CMY-2 gene. This gene, along with bla CTX-M-1 , is the most common ESBL/AmpC βlactamase in poultry E. coli isolates (50). While in most quinolone-resistant isolates a chromosomal mutation in gyrA gene was detected, seven both susceptible or resistant isolates carried qnrS-1. Co-occurence of bla-CMY-2, -59 and qnrS-1 was observed in two isolates from the ST429 cluster and all but one ST23 isolates carried the remaining qnrS-1 genes. One may assume that the afore-mentioned fact that these STs are not commonly associated with human disease does not make them epidemiologically irrelevant, for they may still serve as a source of resistance or virulence determinants in horizontal gene transfer. Indeed, the importance of horizontal gene transfer may be assumed from the detection of a multitude of replicons previously associated with both resistance and virulence gene spread (51)(52)(53)(54)(55).

Conclusions
Despite its limitations due to relatively small number of isolates of completely sequenced isolates, this study could be considered a basic overview of APEC diversity and a delineation of paths that are to be followed in more extensive monitoring of virulent Within preliminary characterization of 95 isolates, a slide agglutination test with four commercial antisera (O1, O8, O18 and O78) was performed according to the manufacturer's instructions (Denka Seiken, Japan). Presence of several selected resistance and virulence-associated genes (additional file 3-table 3) was detected by PCR. After this preliminary characterization 32 strains were selected for whole-genome sequencing. We selected isolates from chicken originating from one producer. To encompass the greatest possible diversity, we exluded isolates from the same individual, the same farm or the same date of isolation, if they showed the identical resistance phenotype and gene profile. 15 5.2 DNA extraction and whole-genome sequencing NucleoSpin Tissue DNA extraction kit (Macherey-Nagel, Germany) following manufacturer's instructions was used to obtain pure DNA. The DNA libraries were prepared with Nextera XT Library preparation kit (Illumina, USA). Finally, Illumina Next-Seq and Mi-Seq platforms were used for the whole-genome sequencing to obtain 2 x 150-bp or 2 x 300-bp pairedend reads, respectively.

Data processing
Adaptor residues and low quality (Q ≤ 20) ends were removed from the reads using The threshold for gene identity was set to 95%. The Clermont typing tool was used to classify isolates into phylogenetic groups (64)(http://clermontyping.iameresearch.center/). In order to investigate the genetic relationships between isolates, genomes were annotated using Prokka v1.13 (65) and the core genome alignment was performed using Roary (66). The core genome alignment was used to determine the single nucleotide polymorphism (SNP) distance using snp-dist (https://github.com/tseemann/snpdists). Phylogenetic tree was constructed using RAxML v8.2.10 using GTR+GAMMA+I model (67). Phylogenetic tree was then visualised via iTOL (68)(https://itol.embl.de/). The raw sequencing data were deposited to GenBank under BioProject PRJNA553636 and corresponding accession numbers to SRA for each sample can be found in Table 4 (additional file 4).

Ethic approval and consent to participate
The samples were collected from diseased or dead animals by practical veterinarians in cooperation with the farm owners and with their consent.

Consent for publication
Not applicable.

Availability of data and materials
The data supporting conclusions of this article are available in the GenBank https://www.ncbi.nlm.nih.gov/bioproject/PRJNA553636/.

Competing interests
The authors declare they have no conflict of interest.