Samples were collected from five commercial dairy farms in Hubei during 2019 and 2020. The number of lactating cows in farms A, B, C, D, and E are 202, 300, 400, 1,200, and 287, respectively. Five farms participated in monthly Dairy Herd Improvement (DHI). Lactating cows in five farms were milked 2 times/day in milking parlors. The clinical symptoms of cows with clinical mastitis (CM) are elevated body temperature, redness, and pain in the udder. The judgment of subclinical mastitis (SCM) refers to the Chinese agricultural industry standard NY/T 2692-2015 (National standards of the people’s Republic of China, 2015) [the number of somatic cell count (SCC) is > 500,000 cells/ml and no visible pathological changes]. Daily care of cows in five farms was provided by the veterinarian; the farms routinely used different antibacterial agents such as ceftiofur, amoxicillin/clavulanic acid, lincomycin, florfenicol, kanamycin, and gentamicin for regular prophylactic and treatment protocols. The sample collection includes animal samples (nipple milk, nipple skin swabs, anal swabs from healthy cows, CM cows, and SCM cows) and environmental samples (bedding, feed, feces, air, drinking water, spraying water, washing water, and milk cup swabs). Lactating cows of parity 3rd to 4th were selected for sample collection. Briefly, milk samples were collected from lactating cows by the milkers after premilking disinfection and the first 3 streams of milk were discarded. Commercial liquid delivery mediums (Haibo Biotech, Qingdao, China) were used to collect nipple skin swabs, anal swabs, and milk cup swabs. All the nipples of each cow were rubbed on the skin with swabs 4 to 5 times from the root to the end of teat after milking. For the anal swab, the swab was inserted into the anus of the cow and rotated for 2 to 3 circles. The swab of the milk cup is collected in milking parlors (5 per farm). Feed, bedding, and feces samples were stored in sterile sampling bags and air samples were collected by natural sedimentation method [suspended sterile sampling tube with 15 ml Trypticase Soy Broth (Haibo Biotech, Qingdao, China) on the column in the barn for 15 to 30 min]. Three sampling points were evenly selected in the barn and the interval between sampling points was not less than 5 m and then mix the samples with equal weight. All the samples were stored at 4°C and transported to the laboratory for bacterial culturing and identification within 24 h. SCC of milk sample was analyzed by DHI Center of Livestock and Poultry Breeding Centre in Hubei.

All the samples were inoculated in 5 ml Trypticase Soy Broth (Haibo Biotech, Qingdao, China) and incubated at 37°C for 18 to 24 h. A loopful of broth culture was streaked onto Columbia Blood Agar and MacConkey Inositol Adonitol Carbenicillin (MIAC) agar (Haibo Biotech, Qingdao, China) and incubated at 37°C for 12 to 18 h. A single pink, slimy, and non-hemolytic suspected colony was picked and purified on MIAC agar (Gao et al., 2010). Purified bacteria were selected for biochemical identification and PCR testing. Gram staining was performed with the Gram Kit (Nanjing Jiancheng Bioengineering Institute, Nanjing, China). Biochemical identification tubes (Haibo Biotech, Qingdao, China) were used for motility tests, indole tests, urease tests, oxidase experiments, and citrate utilization tests (Massé et al., 2020). Further identification and confirmation of the isolates were carried out using PCR of khe and 16s rDNA gene (He et al., 2016; Cheng et al., 2021). Positive clones were sequenced by the Tsingke Biotechnology Corporation Ltd. (Wuhan, China). K. pneumoniae ATCC 700603 (khe positive) and Escherichia coli ATCC 25922 (khe negative) were used as control strains.

Genomic DNA of K. pneumoniae was extracted by boiling method (Kuang, 2015). 3 ml suspended plaque samples were incubated for 15 to 20 min at 99°C. Subsequently, the supernatants were immediately transferred to −20°C for 15 min and centrifuged at 12,000 rpm/min for 15 min at 4°C. The supernatants were diluted by sterile water (1:10) and kept at −20°C for PCR testing.

Capsular polysaccharide serotypes were determined by wzi gene sequencing (Brisse et al., 2013). MLST detection method refers to the K. pneumoniae MLST online database¹. The seven housekeeping genes (gapA, infB, mdh, phoE, pgi, rpoB, and tonB) were amplified by PCR. All the positive products were sequenced by Sanger and sequences were submitted to the website² to get the wzi allele types and the subtypes of each housekeeping gene of MLST. Submit the allele profile to the MLST website³ to get ST. wzi alleles and STs that had not been previously described were submitted to the curator of the PubMLST database and were assigned new designations. Sanger sequencing was completed by Tsingke Biotechnology Corporation Ltd. (Wuhan, China).

The ST phylogeny tree (Tamura 3-parameter model) was constructed using the maximum likelihood method in the MEGA-X software and was modified visually with interactive tree of life (iTOL)⁴. The phylogenetic tree is tested by the bootstrap method and the number of tests is 1,000 times. The information of 726 K. pneumoniae isolates from different countries and 724 K. pneumoniae isolates from different hosts was downloaded from the PubMLST database (the same 100 STs as in this study). Minimum spanning tree (MST) was constructed by Bionumerics 8.1.

All the 239 K. pneumoniae were evaluated for the presence of 23 known virulence genes, including transporter (peg344), fimbriae synthesis-related gene (fimH, mrkD), lipopolysaccharide-related gene (uge, wabG), capsular polysaccharide synthesis and synthesis regulation-related gene (wcaG, _crmpA, _prmpA, _prmpA2, and magA), iron uptake system (iroB, iroN, aerobactin, iutA, irp2, iucA, ybtA, kfu, and entB), urease-related gene (allS, ureA), tellurite resistance gene (terB), and hemolysin (hly) by PCR (Supplementary Table 1).

The susceptibility of K. pneumoniae isolates to 17 antimicrobials, which are commonly used on dairy farms and become clinically important broad-spectrum antimicrobials, was determined using the microbroth dilution method recommended by the Clinical and Laboratory Standardization Institute [Clinical and Laboratory Standards Institute [CLSI] (2021)], including penicillins: ampicillin (AMP); β-lactam/β-lactamase-inhibitor combinations: amoxicillin/clavulanic acid (AMC); cephalosporins: cephalothin (CEP), ceftiofur (EFT), and ceftriaxone (CRO); carbapenems: meropenem (MEM); aminoglycosides: streptomycin (STR), kanamycin (KAN), and gentamicin (GEN); tetracyclines: tetracycline (TET) and doxycycline (DOX); amphenicols: florfenicol (FFC) and polymyxins: colistin (CT), rifamycins: rifaximin (RFX), fluoroquinolones: ciprofloxacin (CIP); and sulfonamides: sulfisoxazole (SOX) and sulfamethoxazole/trimethoprim (SXT). Results were interpreted according to the CLSI M100-Ed31 and VET01S-Ed5 standards [Clinical and Laboratory Standards Institute [CLSI] (2020, 2021)]. No CLSI interpretative criteria for ceftiophene and streptomycin are currently available and the resistance breakpoint of ceftiophene and streptomycin refers to ceftiofur and kanamycin, respectively. The criteria for resistance of rifaximin were analyzed by Ecofinder software recommended by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) (Turnidge et al., 2006). The phenotypic resistance patterns are categorized into MDR, extensively drug-resistant (XDR), and pandrug-resistant (PDR) as previously described by Magiorakos et al. (2012). The quality control strain is K. pneumoniae ATCC 700603 and Escherichia coli ATCC 25922.

Antimicrobial resistance genes that confer resistance to β-lactams (bla_CTX–M–2, bla_CTX–M–10, bla_CTX–M–14, bla_CTX–M–15, bla_SHV, bla_TEM, and bla_OXA–1), carbapenems (IMP, VIM, bla_OXA–48, bla_OXA–181, NDM, and KPC), AmpCs (MOX, CIT, DHA, ACC, EBC, and FOX), aminoglycosides [armA, rmtA, rmtB, rmtC, rmtD, rmtE, npmA, aadA1, aadA2, aadB, aacC1, aacC2, aac(3)-IV, aacA4, aphA1, aphA2, aphA6, strA, and strB], plasmid-mediated quinolone resistance (PMQR) genes [aac(6′)-Ib-cr, qnrA, qnrB, qnrC, qnrD, and qnrS], tetracyclines [tet(A), tet(B), tet(C), tet(D), tet(E), and tet(G)], sulfonamides (sul1, sul2, and sul3), macrolides (mefA, ereA, ereB, ermB, mphA, and mphB), polymyxins (mcr-1 to mcr-5), and phenicols (floR, cmlA, and Cat1) were investigated by PCR (Supplementary Table 2). For each positive gene, some PCR products were selected and sent to Quintara Biotechnology Corporation Ltd. (Wuhan, China) for Sanger sequencing for further confirmation.

All of ciprofloxacin-resistant K. pneumoniae were selected to determine the DNA sequence of quinolone resistance-determining region (QRDR) genes (gyrA, gyrB, parC, and parE) (Horii et al., 2008; Pitondo-Silva et al., 2015; Cheng et al., 2018). Three ciprofloxacin-intermediate and 7 ciprofloxacin-sensitive isolates were randomly selected as controls. Genomic DNA was extracted by the Universal Genomic DNA Kit of ComWin Biotechnology Corporation Ltd. (Beijing, China). PCR products were purified and sequenced in both the directions by Quintara Biotechnology Corporation Ltd. (Wuhan, China). The nucleotide sequences and the deduced amino acids were compared with that of K. pneumoniae ATCC 13883 (GenBank JOOW00000000.1) using the Bioedit (Azargun et al., 2019).

SPSS statistics version 26 and GraphPad Prism version 8.0.1 software were used for statistical analysis and the chi-squared test (χ²) was used to compare the statistical significance between the different groups. p < 0.05 means significant difference and p < 0.01 indicates extremely significant difference. Correlations were assessed by calculating the Spearman’s rank correlation coefficient (r). |r| ≥ 0.8 means high correlation, 0.5 ≤ | r| < 0.8 means moderate correlation, 0.3 ≤ | r| < 0.5 means low correlation, and | r| < 0.3 means no correlation.

Klebsiella pneumoniae isolates grew well on MIAC agar due to resistance to carbenicillin and gave characteristic pink colonies as a result of inositol and adonitol fermentation. On Columbia Blood Agar, the colonies are not hemolytic and tend to be viscous and stringy. K. pneumoniae was observed as Gram-negative bacillus under a light microscope. Besides, it showed the characteristics of non-motility, indole negativity, production of urease, negative oxidase testing, and utilization of citrate.

A total of 239 K. pneumoniae (26.94%) were identified in 887 samples through biochemical identification, khe amplification, and 16s rDNA sequencing (Figure 1 and Supplementary Table 3). The samples collected in farm C (114/338, 33.73%) showed a significantly higher K. pneumoniae prevalence than farm A (18/109, 16.51%), farm B (37/179, 20.67%), and farm E (48/189, 25.40%). The K. pneumoniae prevalence in farm D (22/72, 30.56%) was also significantly higher than farm A (18/109, 16.51%) (p < 0.05). There was no significant difference in the K. pneumoniae prevalence of nipple milk, skin swabs, and anal swabs between healthy and mastitis cows (p > 0.05). Besides, K. pneumoniae was found in a variety of environmental sources, such as milk cup swab, feed, bedding, feces, air, drinking water, and spray water and the separation rate is between 15.79 and 47.83%.

Based on the wzi gene sequences, 239 isolates were classified as 101 different wzi allele types (Figure 2). wzi150-KL163KL27KL46 (40/239, 16.74%) accounted for the most and wzi706, wzi727, wzi728, wzi729, and wzi730 were the new wzi allele. The wzi genotypes of 40 strains were unknown, which might be new alleles. K1 and K2 capsule serotypes commonly found in hvKP were not detected in this study. The results of MLST on these strains showed that 100 different STs were obtained (Figure 2), among which ST2854 (30/239, 12.55%) accounted for the largest proportion, followed by ST5960 (15/239, 6.28%) and ST5915 (13/239, 5.44%) and ST5367, ST5369, ST5370, ST5372, ST5855, ST5874, ST5875, ST5876, ST5914, ST5915, ST5917, ST5918, ST5919, ST5920, ST5947, ST5949, ST5951, ST5952, ST5953, ST5954, ST5955, ST5957, ST5959, ST5960, and ST5961 were the new STs. There are still 30 strains of unknown ST, which might be a new allele. We did not find STs such as ST11 and ST23 that have been reported to have a global epidemic of highly toxic in humans. According to the ST phylogeny tree of 239 isolates (Figure 2), K. pneumoniae isolated from five farms in Hubei has a rich variety of serotypes and genotypes. Strains of the same serotype may have differences in the genome and strains of the same genotype may also have different serotypes, such as ST2584 and ST5370 are wzi150, isolates of ST5960 have at least 10 different wzi types.

The MST of 100 different STs showed that there was diversity in the population structure of STs in this study (Figure 3). The founder ST20 and ST17 and ST22 with single locus variants (SLVs) comprise clonal complex 20 (CC20); the founder ST200 and ST292, ST966, ST1537, and ST5627 with SLVs comprise CC200. K. pneumoniae isolated from milk, nipple swab, feed, and feces is classified in the same clone complex, which indicated that the contamination of milk may have a great correlation with the hygiene of nipple surface and environment.

Compared with the PubMLST database, we found that 58 STs were isolated in other countries (Figure 4A). These genotypes are mainly from Italian isolates; ST17, ST20, ST29, ST34, ST36, ST37, and ST101 have a wide range of the region. Furthermore, the host sources of 67 STs were not only cows, but also humans, pigs, chickens, dogs, cats, horses, rhizosphere, environment, and so on. According to Figure 4B, there are 66 STs known to come from multiple hosts, of which 60 STs had existed in humans.

Sixteen of 23 virulence genes were positively detected by PCR (Figure 2). The detection rate of fimbriae synthesis-related gene (fimH, mrkD), lipopolysaccharide-related gene (uge, wabG), iron uptake system (entB, iutA, and iroN), and urease-related gene (ureA) was over 85% and others were between 0.42 and 35.56%. Notably, there are 3 isolates [ST34 (n = 2), ST2410 (n = 1)] carrying iucA (Figure 2), which are related to high virulence. Besides, 99.16% (237/239) K. pneumoniae carried at least 5 virulence genes and three isolates were positive for 12 virulence genes (Figure 2). There are differences in the virulence genes carried by strains of the same serotype or ST; no significant correlation has been found between the virulence genes and the serotypes or genotypes in this study.

The results of the susceptibility test to 17 antimicrobials are shown in Figure 5 and Supplementary Table 4. K. pneumoniae isolates were highly sensitive to meropenem (99.16%) and colistin (99.58%). However, K. pneumoniae was completely resistant to ampicillin and highly resistant to sulfisoxazole (94.56%); the resistance rate to cephalothin, streptomycin, gentamicin, tetracycline, doxycycline, florfenicol, and sulfamethoxazole/trimethoprim was between 20.08 and 47.28%. It is worth noting that 43.93% (105/239) of the isolates were MDR strains and neither XDR nor PDR strains were found in this study (Figure 6A). The resistance rate of K. pneumoniae to ceftriaxone, streptomycin, and gentamicin in mastitis milk (nipple milk of SCM and CM cows) is significantly higher than that in healthy milk (p < 0.05) (Figure 6B). Ninety-three resistance profiles have been identified and the dominant ones were AMP-SOX (57/239, 23.85%), followed by AMP-CEP-SOX (26/239, 10.88%), AMP-STR-GEN-SOX (11/239, 4.66%), and AMP-CEP-STR-GEN-SOX (11/239, 4.66%) (Supplementary Figure 1).

Thirty of 68 AMR genes were identified by PCR (Figure 7). The detection rate of bla_TEM, bla_SHV, strA, strB, aadA1, and aac(6′)-Ib-cr was more than 50% and others were between 0.42 and 38.08%. Eleven CTX-M-producing K. pneumoniae were found. Among the isolates resistant to β-lactam or aminoglycoside antibiotics, more than 90% of isolates have detected at least one β-lactam resistance gene or aminoglycoside resistance gene; among the isolates resistant to TET, DOX, FFC, CIP, and SXT, more than 60% of isolates have detected at least one resistance gene corresponding to the resistance phenotype (Supplementary Figure 1 and Supplementary Table 5). The correlation analysis performed between different phenotypic antibiotic resistance and the antibiotic resistance genes is shown in Figure 8. Among β-lactam antibiotic phenotypes and antibiotic resistance genes, the obtained results revealed low positive correlations between EFT, CRO, and bla_CTX–M–15 (r = 0.492–0.474) and AMC and bla_OXA–1 (r = 0.590). Moreover, a low positive correlation was observed between CEP and EFT (r = 0.405). Among aminoglycosides, a high positive correlation was found between GEN and aacC2 (r = 0.823); moderate positive correlation was observed between STR and GEN (r = 0.606), STR and strB (r = 0.501), and KAN and aphA1 (r = 0.568); low positive correlation was observed between STR and aacC2 (r = 0.488) and KAN and aacA4 (r = 0.428). Among tetracyclines, moderate positive correlations were observed between TET, DOX, and tetA (r = 0.580–0.686). Among sulfonamides and amphenicols, low positive correlation was observed between sul1 and sul2 (r = 0.482), sul2 and sul3 (r = 0.303), FFC, floR, and cmlA (r = 0.316–0.457). No significant association was found between fluoroquinolone resistance phenotype and PMQR genes. DNA sequencing of QRDR of gyrA, gyrB, parC, and parE demonstrated that no amino acid changes were found in these genes whether K. pneumoniae isolates were resistant, moderately resistant, or sensitive to ciprofloxacin and silent mutations found are shown in Supplementary Table 6.