Between March 2017 and February 2019, 115 minks who died of HP were collected from 21 farms in northern China (Jilin, Liaoning, Heilongjiang, Hebei, and Shandong Provinces) ( Figure 1 ). The minks showed clinical pneumonia signs, including: high fever (>40°C), severe depression and anorexia, dyspnea, coughing, as well as hematemesis of the mouth and nose.The dead minks underwent dissection, and lung samples were used for bacterial isolation.

After necropsy, samples were fixed in 10% neutral-buffered formalin. Then, they were embedded in paraffin wax and stained with hematoxylin and eosin for histopathology following standard procedures (Bancroft and Stevens, 1990).

Isolation and identification of E. coli strains were undertaken according to standard procedures with some modifications (Qiu et al., 2019). Lung samples were inoculated onto MacConkey agar. Bacterial colonies were selected and cultured in Luria–Bertani (LB) broth at 37°C. These bacterial isolates were identified by classical biochemical methods and confirmed by 16S rRNA-sequencing. All E. coli strains were grown in LB broth at 37°C with aeration and stored at −80°C in 20% glycerol for further use.

The genomic DNA of E. coli isolates was extracted using the TIANamp Bacteria DNA Kit according to manufacturer's (Tiangen, Beijing, China) instructions.

Traditional agglutination test and O-genotyping PCR were used to identify the O serotypes of the E. coli isolates. Briefly, serotyping of E. coli isolates was done by agglutination test with specific serum against E. coli O antigens (Statens Serum Institut, Copenhagen, Denmark) according to the manufacturer’s guidelines. The O-genotyping PCR were carried out through the implication of O antigen biosynthesis genes as described previously (Iguchi et al., 2015).

The minimal inhibitory concentration of 85 E. coli strains was determined according to the Performance Standards for Antimicrobial Susceptibility Testing of the Clinical Laboratory Standards Institute (CLSI; M100-S23). Each strain was evaluated for susceptibility to the cefotaxime, tetracycline, chloramphenicol, florfenicol, levofloxacin, ciprofloxacin, sulfamethoxazole, colistin, gentamicin, and spectinomycin. The susceptibility of E. coli isolates was interpreted referring to CLSI M100-S23. E. coli ATCC25922 were used as a quality-control strain in antimicrobial susceptibility testing.

E. coli strains were classified as eight phylogenetic groups using the rapid phylogenetic grouping PCR as described previously. Thus, the phylogenetic groups of E. coli isolates were determined via PCR detection of genes chuA, yjaA, TspE4.C2, arpA, and trpA according to Clermont protocol (Clermont et al., 2013) ( Table 1 ).

All strains were tested for the presence of 24 virulence-associated genes by simplex PCR and multiplex PCR (Tibbetts et al., 2003; Meng et al., 2014): LT, STa, STb, SLT-Ⅰ, cnf-Ⅰ, cnf-Ⅱ, eae, K99, aatA, papC, tsh, fimC, mat, ibeB, vat, yijp, ibeA, ompA, neuC, cva/cvi, iss, fyuA, iucD and hlyF. The primers used for detection are listed in Table 2 . PCRs were conducted under identical reaction conditions according to those described previously (Tibbetts et al., 2003; Meng et al., 2014).The amplification products were visualized by agarose gel (1%) electrophoresis.

Whole-genome sequencing (WGS) was used to characterize the resistome, MLST and evaluate genetic evolution in 66 representative strains. The whole genome was sequenced by Beijing Novogene Bioinformatics Technology (Beijing, China). The sequence of genomic DNA in cells was identified (PacBio Single Molecule, Real-Time Sequencing; Pacific Biosciences, Menlo Park, CA, USA). The neighbor-joining method (>5 samples) was employed to construct an evolutionary tree using TreeBeST Vision 1.9.2. Resistance genes were predicted by comparing the sequenced strains with a database on antiviral resistance genes (http://katholt.github.io/srst2/).

Sixty-six E. coli strains were underwent MLST by comparing of seven housekeeping genes (adk, fumC, gyrB, icd, mdh, purA, and recA) according to WGS results and aligned against the allele templates of E. coli retrieved from an online database (http://www.mlst.net/). A minimum spanning tree and clustering tree were generated using BioNumerics 7.6 (www.applied-maths.com/download/software/).

Biofilm-formation assays were undertaken, as described previously (Vogeleer et al., 2015) with some modifications. Briefly, E. coli strains were cultured in LB overnight at 37°C and diluted into LB at 1:100. A 200-μL aliquot of each dilution was dispensed into individual wells of a sterile 96-well microtiter plate and incubated for 36 h at 37°C. Wells with sterile LB medium served as negative controls. Then, the culture medium was discarded, and the wells were washed thrice with phosphate-buffered saline. After fixation with methanol for 15 min, biofilms were stained with 0.1% crystal violet for 5 min. The plates were washed five times to remove unbound dye and air-dried. Finally, the biofilms were quantified by measuring the optical density (OD) at 595 nm with a microplate reader(eppendorf, BioPhotomete) after dissolving in 33% glacial acetic acid for 30 min. The capability to form a biofilm was scored in accordance with criteria described previously (Stepanovic et al., 2000): OD< ODc, non-producer; ODc < OD < 2×ODc, weak producer; 2× ODc < OD < 4 × ODc, moderate producer; OD >4ODc, strong producer. “OD” denoted the OD at 595 nm of E. coli, whereas “ODc” denoted three standard deviations above the mean OD of the negative control. Tests were carried out thrice, and the results were averaged.

The onsets of minks were breathing difficulties, blood-like exudates around the nostrils/mouths, and, in some cases, foaming around the mouth, The most common observation were edematous, extensive hemorrhage and pleural effusion. hyperemia or hemorrhagic spots and ecchymosis were observed in lung. Other symptoms were liver hemorrhage, renal hemorrhage, spleen enlargement, and black bleeding spots. The intestinal surface and inner wall were dark-red or pink, which indicated a small amount of bleeding. Swelling and bleeding of inguinal, submandibular, and mesenteric lymph nodes were also documented ( Figure 2 ).

The hispathological changes of organs were as follows, lung, severe pulmonary edema, a large number of eosinophilic serous exudation in alveolar cavity; tracheae, cell necrosis, nuclear fragmentation or dissolution in mucosal layer; heart and liver: lymphocytic infiltration around blood vessels; spleen, massive extramedullary hematopoietic foci and multinucleated giant cells; Intestine, no obvious inflammation; kidney, necrotic shedding of renal tubular epithelial cells ( Figure 3 ).

A total of 115 lung samples from 21 mink farms were cultured. Eighty-five E. coli isolates were identified by conventional biochemical methods. 16S rRNA-sequencing indicated that these strains had features consistent with E. coli.

Serotyping showed that 87.1% (74/85) of E. coli isolates could be classified into the single-O type, whereas 11 isolates could not be classified into any serotype. Of these serotyped isolates, 18 O serotypes were identified ( Figure 4 ). The most prevalent serotype were O11 (39/85), followed by O9 (6/85), O25 (6/85), O8 (3/85), O39(3/85), O69(3/85), O78(2/85), and O112ac (2/85). Only one strain belonged to O5, O7, O29, O33, O45, O88, O89, O105, O109, and O177 serotypes, respectively. These results indicated that O11, O9, and O25 were the predominant serotypes of extraintestinal pathogenic E. coli from mink in a particular population in northern China.

The results of the antimicrobial susceptibility test are shown in Table S1 of Supplementary Materials . Eighty-five E. coli strains were sensitive to sulfamethoxazole, but they showed high resistance to tetracycline (76.5%), chloramphenicol (71.8%), ciprofloxacin (63.5%), and florfenicol (52.9%). Resistance to colistin was noted in only 5.9% of E. coli strains. Resistance to the other antibiotics tested was 16.5%–35.3% ( Figure 5 ). We analyzed the susceptibility of E. coli strains to 10 antibiotics, and found that 55 of 85 (64.7%) strains were resistant to ≥3 antibiotics.

All 85 E. coli strains were screened for the presence of 24 virulence genes. yijp and fimiC were detected in 84 (98.8%) strains. iucD, ompA, cnf-Ⅰ, and fyuA were detected in 82 (96.5%), 81 (95.3%), 81 (95.3%), and 78 (91.8%) strains, respectively. mat, hlyF, and ibeB were present in 76 (89.4%), 70 (82.3%), and 69 (81.2%) of strains, respectively. iss and cav/cvi were detected in 49 (57.6%) and 40 (47.1%) of strains, respectively. aatA, ibeA, vat, and eae were detected in 13 (15.3%), 9 (10.6%), 5 (5.9%), and 4 (4.7%) strains, respectively. Moreover, three (3.7%) strains harbored STa and nueC. In general, all E. coli strains contained ≥4 virulence genes. However, LT, STb, SLT-Ⅰ, cnf-Ⅱ, K99, papC, and tsh were not detected ( Figure 6 ).

Phylogenetic typing showed that most of E. coli strains belonged to group E (45.9%, 39/85) and A (27.1%, 23/85). Only 3.5% of E. coli strains belonged to groups B2 and D, 2.4% to F, and 1.2% to clade I. However, two E. coli strains were not classified into any phylogenetic group ( Figure 7 ). Thus, the E. coli strains that caused HP in mink were strongly associated with phylogenetic groups A and E.

Most virulence genes were associated with phylogenetic groups E and A. yijp, fimC, iucD, ompA, and fyuA showed a wide distribution among all groups (A, B1, B2, D, E, F, and clade I). cnf-Ⅰ, hlyF, iss, and cva/cvi were less prevalent in clade I. vat was associated with group B2 and clade I. STa was found only in groups A and E. nueC was associated only with groups A and group B2 ( Table 3 ).

Whole-genome sequencing (WGS) was done on the genome data of the new strains in the present study. The complete genome MLST clustering tree showed that the E. coli strains from mink were far from the reference strains. The E. coli strains in mink showed regional aggregation and transmission. The homology of the E. coli strains from Weihai, Dalian, Shangzhi, Harbin, and Weihe reached 100%, respectively. However, the other strains were far from each other ( Figure 8 ).

We predicted the types and ratios of drug-resistance genes in 66 strains of E. coli by whole-gene sequencing. tet(A), strA, strB, sul2, oqxA, blaTEM-1B, floR, and catA1 had the highest frequency at 50%, 65.8%, 64.5%, 64.5%, 53.9%, 48.7%, 40.8%, and 34.2%, respectively. The other drug-resistance genes were rangely from 1.3%∼29%, ( Figure 9 ). the highest frequency antibiotic genes (ARGs) were resistant to tetracycline, florfenicol, quinolones, chloramphenicol, which were consistent with our drug resistance phenotype.

Sixty-six E. coli strains were classified into 28 sequence types (STs). The dominant types were ST457 (twenty-seven strains from six outbreaks), followed by ST127 (three strains from one outbreak), ST1485 (three strains from one outbreak), ST10005 (three strains from one outbreak), ST58 (two strains from two outbreaks), ST6730 (two strains from two outbreaks), ST93 (two strains from two outbreaks), ST2973 (two strains from two outbreaks), ST162 (two strains from two outbreaks), and ST156 (two strains from one outbreak). The only one strain of ST was ST23, ST58, ST189, ST1638, ST744, ST1140, ST131, ST165, ST4720, ST6831, ST345, ST224, ST48, ST46, ST770, ST7, ST5, and ST12 ( Figure 10 ).

Out of 85 E. coli strains, six (7.1%) possessed strong biofilm-forming ability, 12 (14.1%) possessed moderate biofilm-forming ability, and 64 (75.3%) showed weak biofilm-forming ability. Three E. coli strains could not form biofilms ( Table S1 in Supplementary Materials ).

The E. coli strains isolated from mink were extraintestinal pathogenic and the dominant serotype was O11. A and E were the predominant phylogenetic groups among E. coli strains in the parts of northern China we sampled. E. coli isolated from mink lungs had similar genes, and some strains had the characteristics of regional aggregation and transmission. Although these strains were resistant to multiple drugs, sulfonamides are first-line treatment for HP caused by E. coli in mink in China. Through the study of extraintestinal pathogenic E. coli in mink, we aim to develop a triple vaccine using P. aeruginosa, K. pneumoniae, and extraintestinal pathogenic E. coli against HP in mink.