CRE isolated from fly samples

Among 150 fly samples collected from 12 locations in Dengfeng, 28 samples collected from nine locations were identified as CRE positive: there were 20 CRE-positive samples from the four farms, five from two rural areas, and three from three urban areas (Fig. 1). The CRE-positive rate in the samples collected from farms was higher than that in the samples collected from rural areas and urban areas. For 24 samples, only one CRE strain was isolated. For three samples, two different species of CRE strains were identified in each. In one sample, three different species of CRE strains were isolated simultaneously. In total, 33 CRE strains were isolated: 15 Escherichia coli (45.5%), six Citrobacter freundii (18.2%), two Raoultella ornithinolytica (6.1%), two Moellerella wisconsensis (6.1%), two Proteus mirabilis (6.1%), one Klebsiella oxytoca (3.0%), one Providencia alcalifaciens (3.0%), one Providencia rettgeri (3.0%), one Proteus terrae (3.0%), one Morganella morganii (3.0%), and one Enterobacter cloacae (3.0%).

Fig. 1

Geographical locations of samples and CRE-positive samples. Yellow circles indicate the locations of all collected samples, and red triangles represent the locations of CRE-positive samples

Antimicrobial susceptibility of 33 CRE strains

The 33 CRE strains identified in this study exhibited resistance to first-line clinical antibiotics, including third-generation cephalosporins/carbapenems/colistin. The resistance rate to third-generation cephalosporins was notably high, exceeding 80%. Specifically, the resistance rate was highest for ceftriaxone (93.9%), followed by cefoperazone-sulbactam (84.8%) and ceftazidime (81.8%). The resistance rates to carbapenems, except for meropenem (24.2%), all exceeded 60% [ertapenem (72.7%), imipenem (63.6%)]. Furthermore, 3.0% of CRE isolates were resistant to colistin, and resistance to tigecycline was not detected. For these first-line clinical drugs, 93.9% of CRE isolates were resistant to both third-generation cephalosporins and carbapenems, making clinical treatment difficult. For 15 carbapenem-resistant E. coli strains, 14 strains (93.3%) were also resistant to third-generation cephalosporins and carbapenems. However, all isolates were sensitive to colistin or tigecycline.

The resistance rates to other classes of antibiotics, except for minocycline (18.2%), and nitrofurantoin (18.2%), all exceeded 20%. Some isolates exhibited resistance to aminoglycosides, including gentamicin (69.7%), tobramycin (54.5%), and amikacin (45.5%). Additionally, 93.9% of isolates were resistant to ampicillin-sulbactam; 87.9% of isolates were resistant to cefazolin, cefuroxime, and amoxicillin-clavulanate; 81.8% of isolates were resistant to cefoxitin and tetracycline; 75.8% of isolates were resistant to piperacillin-tazobactam; 72.7% of isolates were resistant to trimethoprim-sulfamethoxazole; 69.7% of isolates were resistant to chloramphenicol; and 60.6% of isolates were resistant to moxifloxacin. In addition, the cefepime and ciprofloxacin resistance rates were nearly 50%. All isolates were MDR, and one strain was resistant to 10 classes of drugs (Fig. 2).

Fig. 2

Antibiotic resistance phenotypes of 33 CRE strains. Orange indicates antibiotic resistance

Genome sequencing of 33 CRE strains

We sequenced all 33 CRE strains, and 63 total plasmid sequences were obtained, which included 53 complete plasmid sequences and 10 nearly complete plasmid sequence fragments. For eight strains, only one plasmid sequence was obtained for each; for 12 strains, two plasmid sequences were obtained simultaneously; for nine strains, three plasmid sequences were obtained simultaneously; and for one strain, four plasmid sequences were obtained simultaneously.

We obtained 34 chromosome sequence fragments, which included 21 complete chromosome sequences and 13 nearly complete chromosome sequence fragments (Supplementary Table S1). Sixteen types of plasmids were identified, and the highest prediction rate was IncFII (22.2%), followed by Col3M (11.1%), IncFIA/IncFIB (9.5%), IncFIB (7.9%), IncX1/IncY (4.8%), IncX4 (4.8%), IncFIA (3.2%), IncHI2/IncHI2A (3.2%), Incl1-l (3.2%), IncFIA/IncFIB/IncFIII (1.6%), IncHI2A (1.6%), IncX1 (1.6%), IncY (1.6%), IncX3 (1.6%), IncQ1 (1.6%), and p0111 (1.6%). Twenty (60.6%) isolates carried more than one plasmid type. Among these plasmid types, the species range of host bacteria carrying IncFII was narrow, including E. coli and C. freundii, and they were mainly isolated from farm 4. In contrast, the species range of host bacteria carrying Col3M was broad, including P. mirabilis, P. rettgeri, M. wisconsensis, P. alcalifaciens, and P. terrae (Fig. 3).

Fig. 3

Relationships among plasmid types, species of host bacteria, and sampling locations for 63 plasmids

There was an average of 16.5 ARGs per isolate, and all isolates carried two or more ARGs. The overall prediction rate for β-lactam resistance genes was 90.1%, among which 66.7% carried blaNDM−1. The predominant phenicol ARG was floR (75.8%). The prediction rate of tetracycline resistance genes was 84.8%, and the predominant genotype was tet(A) (63.6%). The fosfomycin ARG prediction rate was 27.3%. The prediction rate of quinolone resistance genes was 66.7%, and the most prevalent genes were oqxA and oqxB (30.3%). The sulfonamide resistance gene prediction rate was 94.0%, including sul1 (69.7%), sul2 (63.6%), and sul3 (39.4%). The predominant macrolide-resistant gene was mph(A) (45.5%). The trimethoprim resistance gene prediction rate was 70.0%, and the predominant genotype was dfrA1 (30.3%). The main aminoglycoside resistance genes were aph(3’’)-Ib (57.6%) and aph(6)-Id (57.6%). The mcr-1.1 gene (6.1%) was also detected (Fig. 4). Among the 33 isolates, the prediction rate of carbapenem resistance genes was 81.8% (27/33); these genes were isolated from 23 samples, including blaNDM−1 (66.7%) from 20 samples, blaNDM−5 (3.0%) from one sample, blaNDM−9 (6.1%) from two samples, and blaNDM−1+blaIMP−4 (6.1%) from one sample (Table 1). Carbapenem resistance genes were mostly located on plasmids (74.1%), including two blaNDM−9 genes and 18 (63.6%) blaNDM−1 genes. In addition, blaNDM−1, blaIMP−4, and blaNDM−5 were located on chromosomes in six, two, and one strain, respectively.

Fig. 4

Distribution of ARGs in 33 CRE strains. Blue bars indicate the number of ARGs, gray circles indicate the presence of ARGs in various species

Table 1 Carriage rate of carbapenem resistance genes in 33 CRE strains

Spearman correlation coefficients were measured to assess the correlations between every antibiotic resistance phenotype and their related ARGs. As a result, a strong correlation was observed between antibiotic resistance phenotypes (gentamicin, amoxicillin-clavulanate, ampicillin-sulbactam, cefepime, cefoxitin, chloramphenicol, ciprofloxacin, tetracycline, trimethoprim-sulfamethoxazole) and their related ARGs (Fig. 5). However, no significant correlations between carbapenem-resistant phenotypes and carbapenem-related resistance genes were observed, which may be explained by the complex drug resistance mechanism.

Fig. 5

Correlation between antibiotic resistance phenotypes and their related ARGs. The horizontal coordinates indicate antibiotic resistance phenotypes, and the vertical coordinates indicate phenotype-related ARGs. Blue circles represent significant positive correlations. Darker colors represent stronger correlations

blaNDM−1 was the predominant carbapenem resistance gene in our study. In total, 22 strains carrying blaNDM−1 were all resistant to carbapenems, but not every strain was entirely resistant to three types of carbapenem antibiotics (ertapenem, meropenem, and imipenem). Among them, 20 strains carrying blaNDM−1 were resistant to ertapenem, 13 strains were resistant to imipenem, and three strains were resistant to meropenem.

Furthermore, two metrics (recall and precision) were used to assess the prediction ability of blaNDM−1 for carbapenem-resistant phenotypes. The precision values of blaNDM−1 to ertapenem, imipenem, and meropenem resistance phenotypes were 83.3%, 54.2%, and 12.5%, respectively. The recall values of blaNDM−1 to these antibiotics were 83.3%, 61.9%, and 37.5%, respectively. Eight strains were resistant to meropenem in our study, of which three strains carried blaNDM−1; 24 strains were resistant to ertapenem, of which 20 strains carried blaNDM−1; and 21 strains were resistant to imipenem, of which 13 strains carried blaNDM−1. blaNDM−1 was highly correlated with ertapenem resistance and not highly correlated with meropenem resistance. These results demonstrated the blaNDM−1 was highly correlated with ertapenem resistance and not highly correlated with resistance to meropenem.

Clonal transmissions of CRE strains and horizontal transmissions of bla NDM−1

Phylogenetic trees based on SNPs were constructed for 15 E. coli isolates and six C. freundii isolates to identify the clonal clusters of strains. ANI analysis of the plasmids was used to determine the plasmid similarity to assess the clonal proliferation of strains and the horizontal transmissions of ARGs and plasmids. When two strains have similar chromosomes and plasmids, they are both from the same clone. When two strains have comparable plasmids but distinct chromosomes, this is thought to be caused by horizontal transmission of plasmids between strains.

For the phylogenetic tree of E. coli isolates, two strains in clade I were isolated from different samples in farm 2 with nine SNPs. Because they had similar genomes, they may have come from the same clone. In clade II, five strains with the same carbapenem-encoding gene (blaNDM−1) that were isolated from five different samples in the same geographical location belonged to two clusters (516R, 514R and 520R, 518R, 515R). Isolates within the two clusters exhibited 10 and 26 SNPs, respectively, which indicated that two clonal transmission events may have occurred. Strains in clade III were widely distributed, including in rural area 1, rural area 2, farm 1, and farm 3. Three strains (495R_2, 603R_1, and 604R) with the same carbapenem-encoding gene (blaNDM−1) in a cluster with 31 SNPs were isolated from three different samples in two different locations (straight-line distance > 10 km), indicating that there may have been a clonal transmission event (Fig. 6a).

The phylogenetic tree of C. freundii isolates revealed two clades (Fig. 6b). In clade II, four C. freundii strains (607R_1, 610R_1, 608R, and 602R) from farm 3 formed a cluster with only seven SNPs. This indicated that these four strains spread clonally within farm 3.

Fig. 6

Phylogenetic trees of 15 E. coli strains (a) and six C. freundii strains (b). Note The phylogenetic tree is shown on the left, a heatmap of ARG carriage is shown in the middle (blue indicates the presence of ARGs), and the distribution of plasmid types is shown on the right (red indicates the presence of plasmid types)

To further analyze the horizontal transmission of plasmids, we constructed a phylogenetic tree based on ANI analysis for 53 complete plasmid sequences and 10 nearly complete plasmid sequence fragments (Fig. 7a). The plasmids could be divided into two clades, exhibiting significant differences in the carriage rate of blaNDM−1, tet(A), aph(3’)-VI, qnrD1, and rmtB between these clades (P < 0.05). In clade II, a cluster included 12 complete blaNDM−1 IncFII plasmids. Only three recombination-free SNPs were detected in these blaNDM−1 IncFII plasmids, indicating a horizontal transmission event of blaNDM−1 IncFII plasmids occurred among different host bacteria isolated from farm 3, farm 4, and rural area 2 (Fig. 7b).

Fig. 7

Phylogenetic tree of 63 plasmids based on average nucleotide identity (a). Details of SNPs in recombination-free alignments for the red cluster in the phylogenetic trees (b). Phylogenetic tree based on SNPs in the red cluster (c). The phylogenetic tree is shown on the left, a heatmap of ARG carriage is shown on the right (blue indicates the presence of ARGs)

To better clarify the relationships among these 12 plasmid sequences, we constructed a phylogenetic tree based on SNPs (Fig. 7c). Two clusters with no SNPs in recombination-free alignments were identified in these blaNDM−1 IncFII plasmids. In cluster A, four blaNDM−1 IncFII (515R_P1, 518R_P1, 520R_P1, and 513R_P1) plasmids with no SNPs in recombination-free alignments were isolated from four different samples from farm 4 and different host bacteria (E. coli and C. freundii); this provides evidence for cross-host horizontal transmission of the blaNDM−1 IncFII plasmid (Supplementary Figure S1). A horizontal transmission event was also observed in cluster B, in which five blaNDM−1 IncFII plasmids (610R_1_P2, 607R_1_P2, 602R_P2, 608R_P2, and 516R_P2) with no SNPs in recombination-free alignments were isolated from different host bacteria (E. coli and C. freundii) in two different locations (straight-line distance > 28 km). All five blaNDM−1 IncFII plasmids in clade B harbored the ARGs blaNDM−1, blaTEM−1B, aph(3’)-IV, and rmtB; however, one of them lacked the fosfomycin resistance gene fosA, indicating that a gene deletion might have occurred during cross-host transmission (Supplementary Figure S2).