Characterization of NDM-5-producing Enterobacteriaceae isolates from retail grass carp ( Ctenopharyngodon idella) and evidence of bla NDM-5-bearing IncHI2 plasmid transfer between ducks and fish.

We aimed to characterize NDM-5-producing Enterobacteriaceae from aquatic products in Guangzhou, China. A total of 196 intestinal samples of grass carp collected in 2019 were screened for carbapenemase genes. Characterization of bla NDM-5 positive isolates and plasmids was determined by antimicrobial susceptibility testing, conjugation experiments, Illumina HiSeq, and Nanopore sequencing. One Citrobacter freundii and six Escherichia coli strains recovered from seven intestinal samples were verified as bla NDM-5 carriers (3.57%, 7/196). The bla NDM-5 genes were located on the IncX3 ( n=5), IncHI2 ( n=1), or IncHI2-IncF ( n=1) plasmids. All bla NDM-5-bearing plasmids were transferred by conjugation at frequencies of ~10 -4-10 -6. Based on sequence analysis, the IncHI2 plasmid pHNBYF33-1 was similar to other bla NDM-5-carrying IncHI2 plasmids deposited in GenBank from Guangdong ducks. In all IncHI2 plasmids, bla NDM-5 was embedded in a novel transposon, Tn 7051 (IS 3000-ΔIS Aba125-IS 5-ΔIS Aba125- bla NDM-5- ble MBL- trpF- tat-∆ dct-IS 26-∆ umuD-∆IS Kox3-IS 3000), which was identical to the genetic structure surrounding bla NDM-5 found in some IncX3 plasmids. The IncHI2-IncF hybrid plasmid pHNTH9F11-1 was formed by homologous recombination of the bla NDM-5-carrying IncHI2 plasmid and a heavy-metal-resistant IncF plasmid through ∆Tn 1721. To the best of our knowledge, this is the first report on the characterization of bla NDM-5-bearing plasmids in fish in China. The IncHI2 plasmid pHNBYF33-1 may be transmitted from ducks, considering the common duck-fish freshwater aquaculture system in Guangdong. Tn 7051 is likely responsible for the transfer of bla NDM-5 from IncX3 to IncHI2 plasmids in Enterobacteriaceae, resulting in the expansion of transmission vectors of bla NDM-5.

INTRODUCTION

Carbapenemase-producing Enterobacteriaceae (CPE), especially New Delhi metal-β-lactamase (NDM) producers, have been increasingly reported worldwide and pose a significant challenge to public health (Wu et al., 2019). Since the discovery of NDM-1 in 2008 (Yong et al., 2009), NDM-producing Enterobacteriaceae have spread globally. To date, 41 NDM enzyme variants (NDM-1–NDM-41) (https://www.ncbi.nlm.nih.gov/pathogens/refgene/#NDM) have been identified, with theblaNDM-1 and blaNDM-5 genes being the most prevalent (Wu et al., 2019).

In 2011, blaNDM-5 was first reported in an E. coli strain isolated from a patient in the UK (Hornsey et al., 2011). Since then, blaNDM-5 has been detected in more than 40 countries (https://www.ncbi.nlm.nih.gov/pathogens/microbigge/#blaNDM-5). NDM-5 is the most common NDM variant in E. coli, especially in China and Southeast Asia. Although blaNDM-5 is widespread due to diverse self-transferable plasmids such as IncX3 and IncF (FII, FIA, and FIB) (Wu et al., 2019), it is rarely reported in the IncHI2 plasmid, except in swine- and duck-origin E. coli from Guangdong Province, China (Ma et al., 2021; Zhao et al., 2021b).

In China, NDM-5-producing Enterobacteriaceae have been widely detected in humans (Tian et al., 2020), farm animals (Ma et al., 2021), companion animals (Wang et al., 2021a), wild animals (Wang et al., 2017), retail meats (Zhang et al., 2019), and the environment (Zhao et al., 2021a), but rarely in aquatic products. Aquatic products are also considered important reservoirs and transmission vectors of resistant bacteria (Xu et al., 2020). Of note, the integrated duck-fish freshwater aquaculture system is very common in Guangdong, and antimicrobial resistant bacteria can be transmitted between ducks and fish (Shen et al., 2020). Grass carp (Ctenopharyngodon idella) is the most popular freshwater fish in aquaculture and is cultivated in 32 provinces in China. According to the China Fishery Statistical Yearbook 2020 (https://data.cnki.net/trade/Yearbook/Single/N2021020168?z=Z009), the production of grass carp reached 5.5 million tons in 2019, accounting for 21.7% of the maximum annual production in freshwater fish. However, the occurrence of clinically important resistant bacteria, such as CPE, in grass carp has rarely been studied. Hence, we investigated the prevalence of CPE in intestinal samples of grass carp from wet markets in Guangzhou and characterized blaNDM-positive isolates and plasmids to understand the transmission mechanism of blaNDM5 in aquatic products.

MATERIALS AND METHODS

Sample collection, bacterial isolation, and detection of blaNDM

In January 2019, a total of 196 intestinal samples from grass carp were randomly collected from 24 wet markets located in seven districts of Guangzhou, China. We collected fish samples from different stalls, with three intestinal samples randomly collected from each sampling booth. Each sample was placed in a separate sterile sample bag and transported to the laboratory in a freezer box for processing within 12 h. The fish intestines were dissected with sterile surgical scissors, and 1 g of intestinal content was enriched in 2 mL of Luria-Bertani (LB) broth at 37 °C overnight with shaking. The overnight cultures were streaked onto MacConkey agar plates supplemented with 1 mg/L meropenem and incubated at 37 °C for 18–24 h. Enterobacteriaceae colonies with different morphologies were selected from the plates to screen for blaNDM-, blaKPC-, and blaOXA-48-positive isolates using polymerase chain reaction (PCR) with specific primers as described previously (Poirel et al., 2011).

Antimicrobial susceptibility testing

According to the recommendations of the Clinical and Laboratory Standards Institute (2017), the minimal inhibitory concentrations (MICs) of 18 antimicrobials against NDM-positive Enterobacteriaceae isolates were determined using the agar dilution or broth microdilution (colistin and tigecycline) methods. Escherichia coli ATCC 25922 was used as the control. The MICs were interpreted according to the criteria of CLSI (M100-S30) and EUCAST (http://www.eucast.org).

Conjugation experiments

In this study, streptomycin-resistant E. coli C600 was used as the recipient, and each blaNDM-positive isolate was used as the donor for conjugation by broth mating at 37 ℃ for 16–20 h. Transconjugants were selected on MacConkey agar plates supplemented with 3 000 mg/L streptomycin and 1 mg/L meropenem. Conjugation frequency was calculated following previously reported methods (Chen et al., 2007).

Whole-genome sequencing and bioinformatics analysis

Whole-genomic DNA of NDM-positive isolates was sequenced using the Illumina Hiseq X Ten and Oxford Nanopore MinIon platforms, and complete genomes were obtained by hybrid assembly using Unicycler v0.4.7 (Wick et al., 2017). MLST v2.19 (https://github.com/tseemann/mlst) was used to identify the sequence type (ST) of the blaNDM-positive strains. Plasmid replicons, antimicrobial resistance genes, and heavy metal resistance genes were analyzed using ABRicate v1.0 (https://github.com/tseemann/abricate) with the PlasmidFinder (Carattoli et al., 2014), ResFinder (Zankari et al., 2012), and AMRFinderPlus databases (https://github.com/ncbi/amr), respectively. Plasmid double-locus sequence typing (pDLST) for IncHI2 plasmids was identified using pMLST v2.0 (https://cge.cbs.dtu.dk/services/pMLST/). Insertion sequence (IS) elements were identified using ISfinder (https://isfinder.biotoul.fr/). Single nucleotide polymorphism (SNP) calling was performed using Snippy (https://github.com/tseemann/snippy). The blaNDM-carrying plasmids were further compared and analyzed using the BLAST ring image generator (Alikhan et al., 2011). The genetic context of blaNDM was analyzed by GalileoTM AMR (http://galileoamr.arcbio.com/mara/), Gene Construction Kit v4.5 software (Textco BioSoftware, USA), and Easyfig v2.2.5 (http://mjsull.github.io/Easyfig/files.html).

Nucleotide sequence accession numbers

The complete genome sequences of seven blaNDM-5-positive Enterobacteriaceae were deposited in GenBank under BioProject No. PRJNA636005.

RESULTS

Characterization of blaNDM-5-carrying isolates

A total of seven (3.57%) unduplicated carbapenem-resistant isolates, including six E. coli and one C. freundii, were obtained from the seven intestinal samples of grass carp (Table 1). All seven isolates were identified as blaNDM-5-positive by PCR and sequencing, while blaKPC and blaOXA-48 were not detected.

Table 1

Characterization of blaNDM-5-carrying Escherichia coli andCitrobacter freundii isolates

Isolates	Species	MLST	Farmers market (FM)	Other resistance genes	Heavy metal-resistant genes	Chromosomal mutations	Location of blaNDM-5	Plasmid size (bp)	Conjugation frequencya
a: Average±Standard error (SE). N.D.: Not detected. Bold: blaNDM-5−carrying plasmids and other resistance genes.
BY9F33M	E. coli	ST57	FM4	aac(3)-IV, aadA2b, aph(3'')-Ib, aph(3')-Ia, aph(6)-Id, aph(4)-Ia, blaTEM-1B, floR, sul3, cmlA1 aadA1 , aadA5, aph(3')-IIa, mdf(A), mph(A), qnrS1, arr-2, sul1, tet(A), dfrA14, dfrA27	terDZW,merPRT, arsADR	GyrA (p.S83L, p.D87N) ParC (p.S80I)	IncHI2, IncFIB, IncX1, IncY	238 926	(4.93±0.91)×10−6
BY9F36M	C. freundii	ST557	FM4	aac(3)-IId, aadA16,aac(6')-Ib-cr, aph(3'')-Ib, aph(6)-Id, blaTEM-1B, blaCMY-129, mph(A), floR, qnrB18, qnrB26, qnrB6, arr-3, sul1, sul2, tet(A), dfrA27	arsADR, merCPTR	N.D.	IncX3, IncFIB (K)	46 161	(2.05±0.21)×10−4
PY9F04M	E. coli	ST48	FM21	aadA1,aadA2, aph(3')-Ia, blaOXA-10, mdf(A), mph(A), erm(42), cmlA1, floR, oqxAB, arr-2, sul2, tet(A), tet(M), dfrA12, dfrA14	pcoABCDESR,iucABCD, silABCEFPRS	GyrA (p.S83L, p.D87N) ParC (p.S80I)	IncX3, IncFII, IncFIB, IncFIA	46 161	(3.75±0.80)×10−5
PY9F07M	E. coli	ST48	FM21	ant(3'')-Ia, aadA1,aadA2,aph(3')-Ia, blaOXA-10, mdf(A), mph(A), erm(42), cmlA1, floR, oqxAB, arr-2, sul2, tet(A), tet(M), dfrA12, dfrA14	pcoABCDESR,iucABCD, silABCEFPRS	GyrA (p.S83L, p.D87N) ParC (p.S80I)	IncX3, IncFII, IncFIB, IncFIA	46 161	(2.88±1.41)×10−5
PY9F09M	E. coli	ST155	FM21	aadA1,blaOXA-10, mdf(A), cmlA1, floR, qnrS1, arr-2, tet(A), dfrA14	−	N.D.	IncX3, IncFIB	46 161	(4.87±0.25)×10−5
TH9F11M	E. coli	ST101	FM8	aac(3)-IV, aadA2b, aph(3'')-Ib, aph(3')-Ia, aph(6)-Id, aph(4)-Ia, blaOXA-10,blaCTX-M-55, fosA3, cmlA1, floR, qnrS1, sul3, tet(A), dfrA1, dfrA14, arr-3,mdf(A), arr-2, sul2	iroBCDEN, arsADR, terDZW	GyrA (p.S83L, p.D87N) ParC (p.S80I) ParE (p.S458A)	IncHI2-IncF, IncFII, p0111	407 456	(2.79±0.35)×10−5
HZ9F01M	E. coli	ST9124	FM12	aadA22, aph(3')-Ia, aph(6)-Id, blaTEM-1B, mdf(A), mph(A), lnu(F), floR, qnrS1, arr-2, sul3, tet(A), dfrA14	terDZW	GyrA (p.S83L, p.D87N) ParC (p.S80I)	IncX3, IncFII, IncHI2, IncI1	46 161	(2.35±0.31)×10−5

The seven blaNDM-5-carrying isolates showed multidrug-resistant phenotypes and harbored multiple resistance genes (Tables 1, 2). Molecular typing results showed that the C. freundii strain belonged to ST557. Six of the NDM-5-positive E. coli strains belonged to five different STs, namely ST48, ST57, ST101, ST155, and ST9124. The two ST48 E. coli isolates (PY9F04M and PY9F07M) were recovered from the same market but from different sample booths (Table 1) and were related as they showed only 10 core-genome SNP (cgSNP) differences from each other (Schürch et al., 2018), although the resistance genes they carried were not the same (Table 1).

Table 2

Antibiotic susceptibility of blaNDM-5-carrying isolates and their transconjugants

Isolates	Species	MIC (mg/L)
		AMP	FOX	CTX	CAZ	IPM	APR	STR	CIP	DOX	TET	TIG	AMI	GEN	NEO	CL	SXT	FLR	FOS
MIC, minimal inhibitory concentration. AMP, ampicillin; FOX, cefoxitin; CTX, cefotaxime; CAZ, ceftazidime; IPM, imipenem; APR, apramycin; STR, streptomycin; CIP, ciprofloxacin; DOX, doxycycline; TET, tetracycline; TIG, tigecycline; AMI, amikacin; GEN, gentamicin; NEO, neomycin; CL, colistin; SXT, sulfamethoxazole/trimethoprim; FLR, florfenicol; FOS, fosfomycin.
BY9F33M	E. coli	>128	>128	>128	>128	8	>128	128	32	64	>128	0.5	1	32	128	0.25	32	>128	16
BY9F33M-1T		>128	128	32	>128	4	>128	>256	0.25	4	16	0.25	1	8	128	0.125	16	64	16
BY9F36M	C. freundii	>128	>128	>128	>128	8	8	128	8	32	128	0.5	1	32	1	0.25	32	128	16
BY9F36M-1T		>128	>128	64	>128	4	8	>256	0.008	1	1	0.25	1	0.5	1	0.125	0.25	2	16
PY9F04M	E. coli	>128	>128	128	>128	8	8	64	32	64	128	0.5	2	0.5	128	0.25	32	128	16
PY9F04M-1T		>128	>128	64	>128	4	8	>256	0.008	1	1	0.25	1	0.5	1	0.125	0.25	2	16
PY9F07M	E. coli	>128	>128	128	>128	8	8	64	32	64	128	0.5	2	0.5	128	0.25	32	128	16
PY9F07M-1T		>128	>128	128	>128	4	8	>256	0.004	8	1	0.25	2	0.5	1	0.125	0.25	2	16
PY9F09M	E. coli	>128	>128	128	>128	8	8	32	0.25	32	128	0.5	2	0.25	1	0.25	8	128	16
PY9F09M-1T		>128	>128	64	>128	4	8	>256	0.004	1	1	0.25	2	0.25	1	0.125	0.25	2	16
TH9F11-1M	E. coli	>128	>128	>128	>128	8	>128	256	128	64	>128	1	1	32	128	0.125	32	128	>256
TH9F11-1M-1T		>128	>128	128	>128	4	>128	>256	0.25	16	128	0.5	1	8	>128	0.125	16	64	>256
HZ9F01M	E. coli	>128	>128	128	>128	8	8	256	32	64	>128	1	2	0.5	64	0.25	32	128	16
HZ9F01M-1T		>128	>128	64	>128	4	8	>256	0.25	1	64	0.5	2	0.5	1	0.125	0.25	2	16

Characterization of blaNDM-5-bearing plasmids

The conjugation experiments indicated that the seven blaNDM-5-carrying plasmids could be successfully transferred to the recipient E. coli C600 strain, and replicon typing results revealed that the blaNDM-5 genes were located on IncX3 (n=5), IncHI2 (n=1), and IncHI2-IncF (n=1). The conjugation frequencies of the IncX3-type plasmids varied from ~10−4 to 10−5 cells/donor, while the conjugation frequencies of the IncHI2-type and IncHI2-IncF-type plasmids were ~10−6 and ~10−5 cells/donor, respectively (Table 1).

The complete sequences of all seven blaNDM-5-bearing plasmids were obtained using Illumina and Nanopore sequencing. The sequences of five IncX3 plasmids were similar to previously reported blaNDM-5-bearing IncX3 plasmids, including plasmids pGDQ8D112M-NDM (GenBank Accession No. MK628734, duck, China), pNDM5_IncX3 (KU761328.1, Homo sapiens, China), pHNYX638-1 (MK033577, pork, China), and pHN7DH6 (MN276080, dog, China) (Figure 1A).

Figure 1

Comparison of blaNDM-5-carrying plasmids

A: blaNDM-5-harboring IncX3 plasmids in this study with other similar plasmids. Plasmid pHNYX638-1 (MK033577) is reference plasmid. pGDQ8D112M-NDM (MK628734), pNDM5_IncX3 (CP065346), pHN7DH6 (MN276080), pHNBYF36-1, pHNPYF04-1, pHNPYF07-1, pHNPYF09-1, and pHNZF01-1. B: blaNDM-5-harboring IncHI2 plasmid pHNBYF33-1 in this study with other similar plasmids. pG3X16-2-2 (CP038139), pMCR1_025943 (CP027202), pHNGD64-NDM (MW296099), pHNTH9F11-HI2 (this study), GDQ8D151 plasmid1 (JAGFYD010000002), GDQ20D15 plasmid1 (JAGFYB010000003), GFQ9D68 Contig5 (JAGFYC010000005), and pNDM33-1 (MN915011). C: blaNDM-5-harboring IncHI2-IncF24:A-:B1 plasmid pHNTH9F11-1 in this study with other plasmids. GDQ8D151 plasmid1 (JAGFYD010000002), pHNBYF33-1 (this study), pHNGD64-NDM (MW296099), GDQ20D15 plasmid1 (JAGFYB010000003), GFQ9D68 Contig5 (JAGFYC010000005), pNDM33-1 (MN915011), pHNTH9F11-1_IncHI2, pHNTH9F11-1_IncF, pPK8568-156kb (CP080127), pH2332-166 (NC_025175), pCombat11I9-2 (CP021728), and pP2-3T (MG014722). All rings are represented from inside to outside. Red arrows, resistance genes; green arrows, mobile genetic elements; purple arrows, phage resistance system; blue arrows, homologous recombination regions. Red and blue semicircles in C represent pHNTH9F11-1_IncHI2 and pHNTH9F11-1_IncF, respectively.

Plasmid pHNBYF33-1, which belonged to IncHI2-ST3, was 238 926 bp in length with a GC content of 46.30% and carried 12 resistance genes. The BLASTn results indicated that plasmid pHNBYF33-1 exhibited high similarity (≥99.9% identity and ≥93.4% coverage) to four blaNDM-5-carrying IncHI2 plasmids deposited in GenBank, i.e., pNDM33-1 (MN915011) (Zhao et al., 2021b), GFQ9D68 Contig5 (JAGFYC010000005), GDQ8D151 plasmid1 (JAGFYD010000002), and GDQ20D15 plasmid1 (JAGFYB010000003) (Figure 1B). Interestingly, all four plasmids were carried by E. coli strains recovered from ducks in Guangdong, China.

Figure 1

Plasmid pHNTH9F11-1 (IncHI2-IncF) was 407 456 bp in size and had an average GC content of 48.03%. pHNTH9F11-1 harbored three different replicons, including IncHI2, IncFII, and IncFIB. BLASTn analysis showed that pHNTH9F11-1 was a cointegrate plasmid comprised of sequences of IncHI2 (designated as pHNTH9F11-1_IncHI2), harboring blaNDM-5, and IncF24:A-:B1 (designated as pHNTH9F11-1_IncF) (Figure 1C). In addition, the hybrid plasmid pHNTH9F11-1 had 87% nucleotide sequence coverage of the IncHI2-IncFII plasmid pP2-3T (MG014722, swine, China). The sequence of plasmid pHNTH9F11-1_IncHI2 harboring blaNDM-5 was similar (≥99.99% identity and 100% coverage) to that of the blaNDM-5-carrying plasmid pHNGD64-NDM (MW296099) from a swine E. coli strain (Ma et al., 2021). Plasmid pHNTH9F11-1_IncF exhibited similarity (≥99.99% identity and ≥90% coverage) to IncF24:A-:B- plasmid pPK8568-156kb (CP080127, chicken, Pakistan), carrying multiple heavy metal resistance genes (arsABCD and sitABCD) and a phage resistance system (BREX, bacteriophage exclusion system) (Goldfarb et al., 2015). Further analysis revealed that pHNTH9F11-1_IncF and pHNTH9F11-1_IncHI2 were bound by two identical ∆Tn1721 transposons (hp-tetR-tet(A)-eamA) with a length of 5 492 bp, suggesting that the cointegrate plasmid pHNTH9F11-1 was formed by homologous recombination of these two plasmids through ∆Tn1721 (Figure 1C).

Genetic contexts of blaNDM-5 genes in IncX3 and IncHI2

All five blaNDM-5-carrying IncX3 plasmids showed identical genetic contexts (i.e., IS3000-ΔISAba125-IS5-ΔISAba125-blaNDM-5-bleMBL-trpF-tat-IS26-ΔumuD-ISKox3) (Figure 2), similar to that of the classical IncX3 plasmid pHNYX658-1 (Zhang et al., 2019). The genetic contexts of blaNDM-5 in pHNTH9F11-1_IncHI2 and pHNBYF33-1 (IncHI2) were similar to other blaNDM-5-carrying IncHI2 plasmids in GenBank, including pNDM33-1, GDQ8D151 plasmid1, GFQ9D68 Contig5, and GDQ20D15 plasmid1 (Figure 2). In these four IncHI2-type plasmids, theblaNDM-5 gene was identically embedded in a novel composite transposon (IS3000-ΔISAba125-IS5-ΔISAba125-blaNDM-5-bleMBL-trpF-tat-∆dct-IS26-∆umuD-∆ISKox3-IS3000) inserted between IS1 and IS10R of the multidrug resistance region of the IncHI2 plasmid with 5 bp target site duplications (TSDs) (ACTTT). Previous research has found that excision of this transposon from the plasmid pNDM33-1 forms a circular intermediate (Zhao et al., 2021b). Here, we renamed this novel 13 918 bp long transposon as Tn7051 (https://transposon.lstmed.ac.uk/). Comparative analysis demonstrated that Tn7051 shared 99.98% nucleotide sequence similarity (two SNP differences) with the genetic context ofblaNDM-5 in the IncX3 plasmid pHNYX638-1 (MK033577, pork, China), except that ISKox3 in Tn7051 was truncated by a copy of IS3000, creating only 545 bp remains (∆ISKox3) (Figure 2). Interestingly, when comparing the Tn7051 sequence, we found a Tn7051-like structure (14 482 bp) in three hybrid plasmids obtained from swine E. coli isolates in China (Yao et al., 2020), namely p4M9F (IncFIA-IncHI1A-IncHI1B, MN256759), p4M8F (IncHI1-IncY-IncFIA-IncFIB, MN256758), and p4M18F (IncHI1-IncY-IncFIA-IncFIB, MN256757). In the Tn7051-like structure, ISKox3 had more residues (1 108 bp) than that in Tn7051. Furthermore, the Tn7051-like structure exhibited 99.97%–100.00% nucleotide sequence identity (0–3 SNP differences) to the genetic context ofblaNDM-5 in the IncX3 plasmid pHNYX638-1. Given that Tn7051 and Tn7051-like transposons were similar to the genetic context of blaNDM-5 in the IncX3 plasmid pHNYX638-1, we speculated that both Tn7051 and Tn7051-like transposons were likely derived from the IncX3 plasmid.

Figure 2

Genetic contexts of blaNDM-5 gene in IncX3 and IncHI2 plasmids

Regions with ≥99.0% nucleotide sequence identity are shaded blue. Open reading frames (ORFs) are depicted by arrows. Red, green, purple, and gray arrows represent resistance gene, insertion sequence, ISKox3, and hypothetical protein, respectively. Rods represent direct repeat sequences.

Although the IncHI2 plasmids pHNBYF33-1 and pNDM33-1 shared the same Tn7051 insertion site (Figure 2), compared with pNDM33-1, the plasmid pHNBYF33-1 lacked the ∆IS3000-∆IS10-IS26-lnu(F)-aadA2-hp-IS26 segment, which could be readily explained by a deletion event mediated by two copies of IS26 located in the same orientation (Harmer & Hall, 2016). The genetic context of blaNDM-5 in the hybrid plasmid pHNTH9F11-1 was the same as that in plasmid pHNGD64-NDM, and was very similar to pNDM33-1, except for the lack of the IS26-∆umuD-∆ISKox3-IS3000-∆IS10-IS26-lnu(F)-aadA2-hp-IS26-hp-IS26-blaTEM-IS1X unit (Figure 2). This may be due to the deletion of genes mediated by homologous recombination between two copies of IS26 in the same direction (i.e., IS26 in Tn7051 and IS26 upstream of ΔTnEc1), as IS26 located upstream of ΔTnEc1 had only an 8 bp TSD (CTTCTGGT) on one side (Figure 2).

Proposed formation model of genetic contexts of blaNDM-5 in IncHI2 plasmids

Based on detailed sequence analysis, the co-integration mechanism of IS26 (Harmer & Hall, 2016), and the copy-out-paste-in mechanism of composite transposons (Piégu et al., 2015), we proposed a genetic environment formation model of blaNDM-5 in plasmids pHNBYF33-1 and pHNTH9F11-1, as shown in Figure 3. The assumed plasmid evolution process was as follows: IS3000 was inserted into ISKox3 of the IncX3 plasmid (Figure 3A), thus forming the IS3000-ΔISAba125-IS5-ΔISAba125-blaNDM-5-blaMBL-trpF-tat-∆dct-IS26-∆umuD-∆ISkox3-IS3000-∆ISkox3 unit (Figure 3B), hypothesized due to the absence of this unit in GenBank. The two same-orientated copies of IS3000, surrounding blaNDM-5, generated the circular intermediate Tn7051 (Figure 3C), which was further inserted into the region between IS1 and IS10 of the IncHI2 plasmid (Figure 3D) with 5 bp TSDs (ACTTT), resulting in the formation of the blaNDM-5-carrying IncHI2 plasmids (i.e., pNDM33-1, MN915011; GFQ9D68 Contig5, JAGFYC010000005; GDQ8D151 plasmid1, JAGFYD010000002; and GDQ20D15 plasmid1, JAGFYB010000003) (Figure 3E). IS3000, located downstream of ISKox3, was truncated by IS26 and, consequently, the IncHI2 plasmids evolved into the structure shown in Figure 3F (hypothesized due to the absence of a similar structure in GenBank). The IS26 that previously truncated IS3000 was recombined with the IS26 adjacent to blaTEM-1 (purple frame in Figure 3F), leading to the deletion of the ∆IS3000-∆IS10-IS26-lnu(F)-aadA2-hp-IS26 segment. As a result, the structure of the hypothetical plasmids (Figure 3F) entered the blaNDM-5-carrying plasmid pHNBYF33-1 (Figure 3G). Additionally, the two IS26 elements in Figure 3F and Figure 3G (light green frame) integrated, resulting in the formation of the blaNDM-5-carrying plasmid pHNTH9F11-1 (Figure 3H). Therefore, in summary, we speculated that Tn7051 may contribute to the transfer of blaNDM-5 from the IncX3 plasmids to the IncHI2 plasmids, and the genetic contexts of blaNDM-5 on the IncHI2 plasmids in fish were likely derived from plasmids carried by ducks in Guangdong, China.

Figure 3

Proposed formation mechanism of genetic environment of blaNDM-5 in plasmids pHNBYF33-1 and pHNTH9F11-1

Purple, blue, black, green, and red arrows represent IS3000, ISKox3, IS26, other insertion sequence, and resistance gene, respectively.

DISCUSSION

As the most common CPE, blaNDM-positive Enterobacteriaceae have been isolated from seafood and aquatic environments in several countries (Das et al., 2019; Köck et al., 2018). Moreover, blaNDM-positive Enterobacteriaceae have been detected in freshwater fish in Vietnam (Nakayama et al., 2022) and farmed fish in Egypt (Hamza et al., 2020). To the best of our knowledge, however, this is the first report of blaNDM in freshwater fish from China. Of concern, as fish intestines are consumed in Guangdong, NDM-5-positive Enterobacteriaceae in the intestines of retail fish products could spread to humans via the food chain.

In China, IncX3 plasmids are the most common type of plasmid carrying blaNDM-5 (Ma et al., 2020). NDM-5-producing IncX3 plasmids are widespread in environmental, animal, and clinical isolates (Ma et al., 2020), but are rarely reported in Enterobacteriaceae of freshwater fish origin. The similar IncX3 plasmids found in this study further highlight the importance of the epidemic IncX3 plasmid in the spread of theblaNDM-5 gene within the entire ecosystem. IncHI2/ST3 plasmids have been reported to mediate the transfer of various antibiotic resistance genes (ARGs), such as fosA3 (Wang et al., 2020), floR (Cao et al., 2020), blaCTX-M (Lü et al., 2020), and mcr (Long et al., 2019; Zhi et al., 2016), as well as various NDM-type carbapenemase genes, such as blaNDM-1, blaNDM-9, and blaNDM-4 (Liu et al., 2017; Oueslati et al., 2021). However, there are very few reports of blaNDM-5-carrying IncHI2 (Ma et al., 2021; Zhao et al., 2021b). Consequently, we downloaded all available complete genomes (n=5 974; as of 1 September 2021) of Enterobacteriaceae submitted to the NCBI assembly database (https://www.ncbi.nlm.nih.gov/assembly/) and found only fiveblaNDM-5-carrying IncHI2 plasmids (four from ducks and one from swine), all of which were from Guangdong, China. Although we could not trace the location of the grass carp farms and investigate the contamination source of the blaNDM-5-positive Enterobacteriaceae, it is worth noting that the detection rate of the blaNDM gene in duck samples from Guangdong is high (>30%) (Wang et al., 2021b) and integrated duck-fish farming is very common in Guangdong (Shen et al., 2020). In the duck-fish farm model, duck feces are discharged without treatment, and a large number of ARGs or residual agents can directly contaminate the fish ponds, promoting the transmission of ARGs between ducks and fish. Thus, considering the high similarity of the blaNDM-5-bearing IncHI2 plasmids in the fish and ducks, and that theblaNDM-5-bearing IncHI2 plasmid is currently only found in Guangdong, we speculate that the blaNDM-5-bearing IncHI2 plasmids found in Enterobacteriaceae from retail fish may have been derived from duck feces-contaminated fish ponds in Guangdong. As such, greater attention should be paid to the transfer risk of antimicrobial resistant bacteria in integrated duck-fish farming.

Here, pHNTH9F11-1 (IncHI2-IncF) was identified as a hybrid plasmid, formed by homologous recombination through ∆Tn1721. In gram-negative bacteria, the fusion of plasmids mediated by insertion sequences, such as IS26, is rather universal, leading to a plasmid that can encode multiple resistance and hypervirulence genes, thereby posing a considerable threat to human health; for example, the co-integration event mediated by IS26 between the blaNDM-5-bearing IncX3 plasmid and blaCMY-2-bearing IncA/C plasmid (Li et al., 2020). Moreover, the fusion of plasmids can expand the number of replicons and host range of plasmids, accelerating the dissemination of ARGs among various bacterial species (Dolejska et al., 2014; Wong et al., 2017). Of note, this fusion can also enable a non-conjugative plasmid to acquire conjugation ability, thereby facilitating the transmission of resistance genes, e.g., the recombination of non-conjugative mcr-1-carrying P7 phage-like plasmid pD72-mcr1 and conjugative F33:A-:B- plasmid pD72-F33 mediated by IS26, forming cointegrate plasmid pD72C with a conjugation frequency of 8×10-3 cells/donor (He et al., 2019). Hence, the cointegrate plasmid pHNTH9F11-1 with multidrug resistance, heavy metal resistance, and phage resistance system (ability to resist invasion of bacteriophages) may provide an advantage for the host to survive in the environment.

Composite transposons can mediate the jump of ARGs between different DNA molecules. The novel Tn7051 and Tn7051-like transposons can both be moved by a copy-out-paste-in mechanism utilizing a double-stranded circular DNA intermediate (Yao et al., 2020; Zhao et al., 2021b), thereby contributing to the transfer of theblaNDM-5 gene and expanding its transmission vectors. It has been widely reported that blaNDM-5 genes are mainly located on narrow-host-range plasmids (e.g., IncX3, IncF, and IncB/O) (Wu et al., 2019). However, the transfer of blaNDM-5 to the IncHI2 plasmid mediated by Tn7051 and to the IncHI1-IncY-IncFIA-IncFIB plasmid mediated by Tn7051-like suggested that these transposons may further accelerate the horizontal spread of the blaNDM-5 gene to various strains and plasmids, like Tn3000 and Tn125, which mediate the between-plasmid jumps of blaNDM-1 and accelerate the transfer of blaNDM-1 in different strains (Acman et al., 2021).

CONCLUSIONS

This study revealed the emergence ofblaNDM-5 in Enterobacteriaceae of fish origin in China. To the best of our knowledge, this is the first report of theblaNDM-5 gene, as well as blaNDM-5-bearing plasmids, in isolates from fish products in China. Our findings indicated that blaNDM-5 in the IncHI2 plasmids may originate from the IncX3 plasmid, transferred by the novel composite transposon Tn7051. Furthermore, the blaNDM-5-bearing IncHI2 plasmid may be transmitted from ducks, considering the common duck-fish freshwater aquaculture system in Guangdong. Based on the concept of “One health”, the surveillance of antibiotic resistance in aquatic products should be strengthened, and more measures should be taken to reduce the transfer of clinically important resistant bacteria, such as CPE, between food-producing animals and animal products.

DATA AVAILABILITY

The datasets in this study can be found in NCBI under BioProjectID PRJNA636005. The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2021) in the National Genomics Data Center (Nucleic Acids Res 2021), China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA005844), publicly accessible at https://ngdc.cncb.ac.cn/gsa.

COMPETING INTERESTS

The authors declare that they have no competing interests.

AUTHORS’ CONTRIBUTIONS

J.H.L. and L.C.L. conceived the research. X.G., Y.Y.L., M.Y.G., K.B.M., W.Y.H., and L.C.L. collected the data. L.C.L., J.H.L., Y.Y.L., X.G., and W.Y.H. analyzed and interpreted the data. Y.Y.L. and L.C.L. drafted the manuscript, J.H.L., W.Y.H., and X.G. revised the report. All authors read and approved the final version of the manuscript.