This study was approved by the Ethics Committee of the Municipal Hospital of Taizhou University, Zhejiang, China, and written informed consent was obtained from each of the participants in accordance with the Declaration of Helsinki. The rights of the research subjects were protected throughout, and we confirm that this study was conducted in our hospital.

The use of human specimens and all related experimental protocols was approved by the Committee on Human Research of the indicated institutions, and was carried out in accordance with approved guidelines.

Four of seventy-one non-redundant multidrug resistant Enterobacteriaceae strains, including Klebsiella pneumoniae Kpn761 (harboring In62), non-typhoidal Salmonella strains Nsa243 and Nsa292 (In37 and In1086, respectively), and Enterobacter cloacae Eco336 (In1085), were recovered from hospitalized patients with clinical infections. The isolates were collected between June 2013 and July 2015, and were primarily assessed for the presence of integrons by antimicrobial susceptibility testing. Bacterial species were identified by 16S rRNA gene sequencing as described previously (Frank et al., 2008). The strains harboring integrons were studied in the study. Escherichia coli TOP10 cells (Invitrogen, Grand Island, NY, United States) were used as a host for cloning experiments, and as a control for susceptibility testing.

The minimum inhibitory concentration (MIC) values of isolates Eco336 and Nsa292 (harboring In1085 and In1086, respectively) for 12 antimicrobial agents, including cephalosporins (cefazolin, ceftazidime, and ceftriaxone), aminoglycosides (netilmicin, tobramycin, and amikacin), carbapenems (ertapenem, meropenem, and imipenem), and quinolones (norfloxacin, ofloxacin, and ciprofloxacin), were determined using the Microscan broth dilution method (Microscan, Renton, WA, United States). The MICs were interpreted in accordance with the Clinical and Laboratory Standards Institute guidelines (Clinical and Laboratory Standards Institute [CLSI], 2015).

Plasmids from four integron-harboring strains were isolated using an AxyPrep Plasmid Miniprep kit (Axygen Biosciences, Beijing, China) according to the manufacturer’s instructions and as described previously (Wang et al., 2014). Two plasmids containing novel integrons In1085 and In1086 were then digested with BamHI (TaKaRa, Dalian, China) and subjected to agarose gel electrophoresis to generate genetic maps.

To characterize the two novel integrons, the relevant BamHI fragments of the two digested plasmids were ligated into the pMD19-T cloning vector (TaKaRa, Dalian, China). The ligation mixtures were electroporated into E. coli TOP10 cells, which were then plated on LB medium with medium supplemented with ampicillin (100 μg/mL) and incubated over-night at 37°C. Plasmids from any resulting aacA4-positive colonies were isolated by PCR amplification and sequencing (Supplementary Table S1), and the inserts were amplified using the primers specified in Supplementary Table S1 and the following thermal cycler conditions: 3 min at 94°C, 30 cycles of 1 min each at 94, 54, and 72°C, followed by 10 min at 72°C. The total reaction volume was 25 μL, and the total eluent volume was 10 μL. Following amplification, the amplicons were separated by gel electrophoresis on a 0.6% agarose gel run at 90 V for 90 min in 0.5 × TBE buffer. Fragments corresponding to plasmid DNA and integrons were recovered, and the initial positions of the relevant genes in the recombinant plasmids were determined according to a previously established method for estimating plasmid DNA sizes (Wang et al., 2003). Thus, the sequences of both integrons, including upstream and downstream genes, were obtained, by PCR amplification and sequencing (Supplementary Table S1), and the genetic structures were mapped and characterized by next generation sequence annotation and genome comparison. Plasmids harboring In1085 and In1086, In1085-TOP10 and In1086-TOP10 were studied, and the values of relevent susceptibility testing were confirmed (Table 1) by conjugation experiments in according with previous report (Wang et al., 2014).

Open reading frames (ORFs) were predicted using RAST¹, and further annotated using BLASTP and BLASTN searches² against the UniProtKB/Swiss-Prot³, and National Center for Biotechnology Information non-redundant⁴ databases. Annotation of drug resistance genes, mobile elements, and other genes was based on CARD analysis⁵, the β-lactamase database⁶, ISfinder⁷, and INTEGRALL⁸. Sequence comparisons were performed using BLASTN and CLUSTALW2⁹. Gene organization diagrams were drawn using Inkscape¹⁰.

The presence of classes A, B, and D carbapenemase activity in cell extracts from non-typhoidal Salmonella strain Nsa292 and E. cloacae strain Eco336 was determined using a modified CarbaNP test (Chen et al., 2015). Briefly, overnight cultures of each strain grown in Mueller-Hinton (MH) broth were diluted 1:100 into 3 mL of fresh MH broth, and then incubated at 37°C with shaking at 200 rpm to an optical density at 200 nm of 1.0–1.4. When required, ampicillin was used at 200 μg/mL. Bacterial cells were harvested from 2 mL of culture, and cell pellets were washed twice with 20 mM Tris-HCl (pH 7.8). Each cell pellet was resuspended in 500 μL of 20 mM Tris-HCl (pH 7.8), lysed by sonication, and centrifuged at 10,000 × g for 5 min at 4°C. Each 50-μL supernatant (containing the enzymatic bacterial suspension fraction) was mixed with 50 μL of the following substrates (I to V), followed by incubation at 37°C for a maximum of 2 h: substrate I: 0.054% phenol red and 0.1 mM ZnSO₄ (pH 7.8); substrate II: 0.054% phenol red, 0.1 mM ZnSO₄ (pH 7.8), and 0.6 mg/μL imipenem; substrate III: 0.054% phenol red, 0.1 mM ZnSO₄ (pH 7.8), 0.6 mg/μL imipenem, and 0.8 mg/μL tazobactam; substrate IV: 0.054% phenol red, 0.1 mM ZnSO₄ (pH 7.8), 0.6 mg/μL imipenem, and 3 mM EDTA (pH 7.8); and substrate V: 0.054% phenol red, 0.1 mM ZnSO₄ (pH 7.8), 0.6 mg/μL imipenem, 0.8 mg/μL tazobactam, and 3 mM EDTA (pH 7.8).

The sequences of novel integrons In1085 and In1086 were deposited in the GenBank database under accession numbers KP870111 and KP870112, respectively.

Four strains, including K. pneumoniae strain Kpn761 (harboring In62), non-typhoidal Salmonella strains Nsa292 and Nsa243 (In1086 and In37), and E. cloacae strain Eco336 (In1085), were selected for further study. Non-typhoidal Salmonella Nsa292 and E. cloacae Eco336 showed resistance to aminoglycoside, cephalosporin, and carbapenem antibiotics (Table 1), while the remaining two strains were only used for reference in this study and will be described elsewhere. E. cloacae Eco336 was isolated from the urine of a urological surgery patient, K. pneumoniae Kpn761 (In62) was isolated from a blood culture of an intensive care unit patient, and non-typhoidal Salmonella strains Nsa292 and Nsa243 (In37) were isolated from blood cultures of patients hospitalized with infections.

Following BamHI digestion and ligation into a pMD19-T cloning vector, the four integron-containing recombinant plasmids were transformed into E. coli TOP10 cells. Antibiotic susceptibility test results for the resulting TOP10 cells containing the recombinant plasmids are listed in Table 1. The recombinant plasmids were also electrophoresed to estimate their sizes. The susceptibility test results also indicated that the conjugation experiments were successful (Table 1), and that the observed antibiotic resistance was conferred by plasmid-mediated genes. The electrophoresis results following BamHI digestion indicated that the sizes of the In37, In62, In1085, and In1086 integrons were approximately 5.3, 5.3, 5.7, and 6.6 kb, respectively. Usually, class 1 integrons that integrate within transposons or are encoded on plasmids display regular mobilization and transformation capabilities (Carattoli, 2003; Mazel, 2006). As such, integrons can change from one type to another, with the possibility of generating novel types (Figure 1).

Both transformant strains containing the novel integrons, In1085-TOP10 and In1086-TOP10, demonstrated class D carbapenemase activity (data not shown), showing resistance to cephalosporins, carbapenems, and aminoglycosides. However, both strains were susceptible to fluoroquinolones (Table 1). Of the four integrons, In62 apparently represents the most primitive form. It carried two different resistance markers, but only one single-gene GC (GCaacA4cr) (Figure 1). In contrast, In37, In1085, and In1086 contained four or five GCs (Figure 1), containing determinants of at least three different classes of antibiotics. These three larger integrons most likely confer the multidrug resistance phenotype that was demonstrated in the results of antibiotic susceptibility testing of both In1085 and In1086. Interestingly, In37 was almost identical to novel integron In1085, except for the lack of GCdrfA27, which was replaced by non-functional remnants of drfA27 and IS26 in In37 (Figure 1). In addition, a hybrid attC, attC_aar-3/catB3, was identified downstream of the GCarr-3 in In1085. This type of hybrid attC is unlikely to be functional for recombination of the GC (GCarr-3). And, instead of the usual 5′-untranslated region (UTR) upstream of the GCdfrA27 in In1085, the upstream attC matched the 5′-UTR of GCarr-3, while the downstream attC was deleted. Therefore, neither GCdfrA27 nor GCarr-3 in In1085 appear to show any integron features, and are likely to be pseudo-integrons. However, the remaining three GCs (GCaacA4cr, GCbla_OXA-1, and GCcatB3) do appear to be functional.

In1086 differed from In1085 by the presence of an extra GC, GCaadA16 (Figure 1). Noticeably, GCaadA16 appeared to be truncated, and was thus named GCaadA16Δ. The attC of GCaadA16 also appeared to be mutated to attC_qacE/aadA16, rather than the normal attC_aadA16 sequence. However, the mutation did not result in any changes in biochemical properties. Thus, In1086 was characterized as containing six GCs. Neither of the novel integrons contained Pc variant sites or 19-bp ORF11 duplicates, but did contain non-functional remnants of intI1 ORFs.

Sequence comparisons revealed a high degree of homology at nucleotide level between Tn6308 (Sun et al., 2016) and pNSA292 (harboring In1085), and between pSA1-like (Soler Bistué et al., 2006) and pECO336 (harboring In1086) (Figure 2). Even greater homology was observed between pNSA292 and pECO336 and pKPS30 (Compain et al., 2014) (Figure 2). pNSA292 shared identity with Tn6308 in both integron regions, except for a lack of GCdfrA27 and GCΔaadA16 in Tn6308, and was confirmed by the high degree of sequence identity between the backbones of Tn6308 and pNSA292 upstream of both integron regions, and at the sap module (Figure 2). pECO336 also showed partial homology to pSA1-like in the integron regions, and to the Tn6309 backbone and sap module in the up- and down-stream integron regions (Figure 2). However, both pNSA292 and pECO336 showed greater homology to pKPS30 in both integron regions (Figure 2).

However, there were a few structural differences between pNSA292 and pECO336, and thus, the plasmids showed different biochemical characteristics. Because of the mobility of transposons and plasmids, we hypothesize that novel integrons In1085 and In1086, located on pEco336 and pNsa292, respectively, have the potential to mobilize. In addition, these integrons may be related to other plasmids and transposons such as Tn6308, Tn6309, Tn1696, pKPS30, and pSA1-like (Figure 2).

Recombination at specific intI1 sites is highly regulated, thereby differing from other reactions mediated by tyrosine recombinases, while recombination at attC sites is much more complicated and variable (Partridge et al., 2009). attC_aadA16 comprises two simple sites, each consisting of a pair of conserved “core sites” (7 or 8 bp), designated R″/L″ and L′/R′ (Recchia and Sherratt, 2002; Escudero et al., 2016). R″/L″ sites are separated by a 7-bp spacer, and L′/R′ by a 7- or 8-bp spacer (Figure 3A). R″/L″ and L′/R′ pairs are separated by a central region (Figure 3A). Usually, R″/L″ and L′/R′ are reverse complements of each other, with R″/R′ and L″/L′ generally being complementary, except for the removal of an extra base in the L″ site [G for a bs array and C for a top strand (ts) array] and two bases in the R′ site (A and C for a bs array and T and G for a ts array) (Figure 3B, in bold and italic). The central region is usually a defective inverted repeat, unless there has been a deletion of a single T (bs array) or A (ts array) nucleotide (Figure 3B, in bold and italic). There are two regions in class 1 integrons: stable arrays, including intI1 and the beginning of the attI1 sequence, and variable arrays, involving the end of the attI1 sequence and the GCs. GC integrations take place at attI × attC recombination sites (Figure 3C), and it is essential that the complete coding sequence of the GC is expressed from the promoter, Pc (Figure 3C). If the GC is inserted in the opposite orientation, it will not be expressed (Figure 3C). Only a few integrons have been shown to contain the Pc promoter (Bissonnette et al., 1991; Tolmasky and Crosa, 1993; Biskri and Mazel, 2003; Szekeres et al., 2007; da Fonseca and Vicente, 2012; Guerout et al., 2013), with most GCs, including the stable arrays within novel integrons In1085 and In1086, appearing to be promoterless (Figures 2, 3). Integration of GCs in the opposite orientation would likely hamper the rapid adaptation of integrons (Nivina et al., 2016). Correct orientation of the GCs confirms the selectivity of the integrase toward the bs (Figure 3C) (Nivina et al., 2016). Functional integration does not follow ts recombination, especially if the ts of attC is presented as a substrate (Nivina et al., 2016), because the attC sequence is not conserved, and the interaction between IntI and the folded attC is mostly non-specific (MacDonald et al., 2006). Hence, specific recognition of the bs is the main determinant of successful GC integration. Therefore, determinants required for specific recognition of the bs are likely to be present in the structural features of the variable GC array (Figure 3C). These results can help in determining the formation of integrons from clinical sources.