SXT/R391 integrative and conjugative elements in Proteus species reveal abundant genetic diversity and multidrug resistance

SXT/R391 integrative and conjugative elements (ICEs) are self-transmissible mobile genetic elements that are found in most members of Enterobacteriaceae. Here, we determined fifteen SXT/R391 ICEs carried by Proteus isolates from food (4.2%) and diarrhoea patients (17.3%). BLASTn searches against GenBank showed that the fifteen SXT/R391 ICEs were closely related to that from different Enterobacteriaceae species, including Proteus mirabilis. Using core gene phylogenetic analysis, the fifteen SXT/R391 ICEs were grouped into six distinct clusters, including a dominant cluster and three clusters that have not been previously reported in Proteus isolates. The SXT/R391 ICEs shared a common structure with a set of conserved genes, five hotspots and two variable regions, which contained more foreign genes, including drug-resistance genes. Notably, a class A β-lactamase gene was identified in nine SXT/R391 ICEs. Collectively, the ICE-carrying isolates carried resistance genes for 20 tested drugs. Six isolates were resistant to chloramphenicol, kanamycin, streptomycin, trimethoprim-sulfamethoxazole, sulfisoxazole and tetracycline, which are drug resistances commonly encoded by ICEs. Our results demonstrate abundant genetic diversity and multidrug resistance of the SXT/R391 ICEs carried by Proteus isolates, which may have significance for public health. It is therefore necessary to continuously monitor the antimicrobial resistance and related mobile elements among Proteus isolates.


Extraction, assembly and annotation of the SXT/R391 ICEs. The contigs of the fifteen SXT/R391
ICEs were extracted and assembled from whole sequenced genomes using SOAP denovo v2.0.4 19 against the reference SXT/R391 ICEs in the genome of P. mirabilis HI4320 (accession number: AM942759.1). The relationships between contigs were displayed using ContigScape 20 with custom primer walking. Sanger sequencing was used to close the gaps in the ICE region and the results were verified by PCR. The Phred/Phrap/Consed software was used for primer design, genome assembly, editing and quality assessment (http://www.phrap.org/consed/consed. html). Regions with low quality of the genome were resequenced. Putative functions were inferred using the Basic Local Alignment Search Tool (BLAST) (http://ncbi.nlm.nih.gov/BLAST) and the ORF finder (http://www.ncbi. nlm.nih.gov/projects/gorf). The RAST (Rapid Annotation using Subsystem Technology, version 4.0) 21-23 server pipeline was used to predict open reading frames (ORFs) and annotate the ORFs of the recovered ICEs.
Phylogenetic and structural analysis of the SXT/R391 ICEs. To investigate the evolutionary origins of the SXT/R391 ICEs in Proteus species, we performed a phylogenetic analysis. First, we used each of the fifteen SXT/ R391 ICEs in a BLASTn search to obtain all high homology ICEs in the public database (last updated in Nov 2015). Then, we scanned and selected 24 ICEs as references, representative for the different evolutionary origins and species based on different scores and identifications (Supplementary Table S1). To construct the phylogenetic tree, core genes of all ICEs were identified using the OrthoMCL software. The concatenated sequences of these core genes were used for phylogenetic analysis. The bootstrap values were calculated based on 1000 replicates. The structures of the fifteen SXT/R391 ICEs were constructed by comparison with the reference ICE in the genome of P. mirabilis HI4320 (AM942759.1) using BLAST and SOAPdenovo 2.04 19 . The tested ICEs were compared against the reference ICE from P. mirabilis HI4320 using Pheatmap (R package version 0.7.7) to plot heatmaps of the annotated core and pan genes of each ICE.

Conjugation experiments.
A mating assay was employed to test ICE mobility. The experiment was conducted as previously described 24 . Transconjugants were selected on LB agar plates containing streptomycin (100 μ g/ml) and kanamycin (100 μ g/ml). The ICE transfer frequency was expressed as the number of transconjugants observed per recipient cell (E. coli SM10). Transconjugants were confirmed by PCR detection of int SXT and antibiotic resistance genes specific for each of the SXT/R391 ICEs. The primer information is listed in Supplementary Table S2.
Sequence data access. The reads sequences and annotated genes of the fourteen ICEs (except for ICEPmiCHN3277) in Proteus species were submitted to GenBank under accession number KX243403-KX243416.

Results
Distribution and general features of SXT/R391 ICEs among Proteus strains. Fifteen out of 123 strains (12.2%, fourteen P. mirabilis and one P. vulgaris strain) were positive for the int SXT gene, including 4.2% (2/48) of food samples (crab and pork), and 17.3% (13/75) of stool samples from the diarrhoea patients, respectively. The genomes of the 15 int SXT -positive strains were obtained and all SXT/R391 ICEs were successfully assembled with the exception of the ICE from isolate TJ3277. The general genomic features of the fifteen ICEs are summarized in Table 1. The SXT/R391 ICEs were designated according to the nomenclature proposed for this family of elements 2 . The lengths of the 15 ICEs ranged from 76,218 bp (ICEPmiCHN1809) to 108,335 bp (ICEPmiCHN3300), with an average length of 93,396 bp. The G + C content ranged from 44.6% (ICEPmiCHN1586) to 47.8% (ICEPmiCHN905/ICEPmiCHN3335) among the SXT/R391 ICEs, with a mean of 46.9%. The total numbers of predicted CDSs were between 53 and 99.
Blast searches of closely related SXT/R391 ICEs in GenBank. As shown in Table 2, the 15 SXT/R391 ICEs were closely related to many different ICEs based on some indicators, such as the highest max score, query coverage and identity. These ICEs derived from Alteromonas macleodii, Alteromonas mediterranea, Vibrio cholerae, Vibrio alginolyticus, Proteus mirabilis and Providencia stuartii. Due to the closely related scores among the hits of each ICE, we listed the top three alignment results for each ICE; the max score ranged from 19,287 to 1.36E + 05, the total score (bits) ranged from 73,099 to 1.95E + 05, the highest query ranged from 46% to 99% and the identity ranged from 96% to100%. Notably, ICEPmiCHN3335 (host strain TJ3335, isolated from the stool of a patient from Tianjin city, see Table 1) was closely related to ICEPmiChn1 (host strain PM13C04, isolated from a chicken in Hubei, China, in 2013, see Supplementary Table S1) with a highest query and identity of 99%. However, some ICEs, such as ICEPmiCHN2407, ICEPmiCHN2410 and ICEPmiCHN2416, only exhibited query coverage from 46% to 67% and identity from 96% to 99%.
Phylogenetic analysis of ICEs in Proteus isolates. The fifteen SXT/R391 ICEs were grouped into six clusters based on the core gene phylogenetic analysis (Fig. 1). Six ICEs, their contained strains isolated from the stool of diarrhoeal patients in two cities (Table 1), belonged to cluster I, with the reference ICEPmiChn1, which was an ICE contained in a P. mirabilis strain isolated from a chicken in Hubei, China (Supplementary Table S1). Three ICEs (strains isolated from food and stool samples from diarrhoea patients in two cities in China) belonged to cluster II, with references including ICEs from V. cholerae strain ICDC-4210 (isolated from the stool of a patient in Jiangxi, China) and P. mirabilis strain HI4320. Cluster III contained ICEPvuCHN2213 from a P. vulgaris strain isolated from food that was closely related to the reference ICE of Alteromonas macleodii MED64 (from waters off the Aegean Sea) 25 . The remaining five ICEs were grouped into three distinct clusters (IV, V and VI) with no references; thus, these ICEs were designated as novel ICEs.
Sequence structure and variable region characteristics of the SXT/R391 ICEs. Except for SXT/ R391 ICE from TJ3277, which was partially assembled, the remaining fourteen SXT/R391 ICEs shared a common structure that was identical to most SXT/R391 ICEs. In addition to the common structure, all fourteen ICEs contained five hotspots (HS1-5) and two variable (III and IV) regions (Fig. 2). VRIII, inserted into rumB gene, was found in all 14 SXT/R391 ICEs. Five ICEs (ICEPmiCHN901, ICEPmiCHN902, ICEPmiCHN903 ICEPmiCHN1586 and ICEPmiCHN3300), contained the dhfR, floR, strB/A and sul2 genes in this region, which conferred resistance to SXT, CHL, STR and SUL, respectively. The tetA and tetR tetracycline resistance genes were present in VRIII of ICEPmiCHN3335 and HS4 of ICEPmiCHN904 and ICEPmiCHN904. Notably, nine ICEs contained a β -lactamase gene in VRIII (Fig. 2 and Supplementary Table S3), which showed 100% identity to the class A β -lactamase gene bla (HMS-1) contained in plasmid R997 from P. mirabilis (GenBank: KX228735) and V. parahaemolyticus strain UCM-V493 (CP007004.1). HS1, HS2, HS4 and HS5 were detected in all ICEs. However, HS3 was detected in only five ICEs with few inserted genes; a dhfR gene was located in HS3 of ICEPmiCHN3300. Abundant foreign genes were inserted into HS4 and HS5, including genes encoding the restriction-modification system (RM) of which the Type I RM was common, serine protease, ATPase, helicase, and exonuclease. Additionally, genes encoding hypothetical proteins with unknown functions were frequently inserted into these two regions. In VRIV, a mer operon was found in ICEPmiCHN2407, ICEPmiCHN2410 and Scientific RepoRts | 6:37372 | DOI: 10.1038/srep37372 ICEPmiCHN2416, this mer operon was also reported at ICEs in bacterial strains from aquaculture environments, previously described for R391 ICEs and mediated resistance to mercury 13 . The annotated genes from all fifteen ICEs and the six closely related ICEs (Supplementary Table S1) were used to construct a heatmap using P. mirabilis HI4320 (AM942759.1) as the reference; ICEPmiCHN3277 was excluded because it was incompletely assembled. The remaining fourteen ICEs shared minor numbers of common genes ( Supplementary Fig. S1) between each of their ICEs and the closely related ICEs (e.g., ICEPmiCHN1809 vs the ICE of Providencia stuartii strain ATCC33672, ICEPmiCHN1586 vs ICEVchCHN2605, and ICEPmiCHN3237 and ICEPmiCHN3277 vs ICEVchCHN4210 had extensive diversity in their annotated genes).
Antibiotic resistance of the Proteus isolates and their relationship with ICEs. Phenotypically, all fifteen SXT/R392 ICE-harbouring isolates presented multi-drug resistance to the 20 tested drugs (Table 1). Isolates MD20140901, MD20140904 and MD20140905 with the most antibiotic phenotypes were resistant to 16 out of 20 drugs. Even the least resistant isolate (TJ3300) was resistant to five drugs. 09MAS2407 and 09MAS2410, MD20140901 and MD20140905, and TJ3237 and TJ3277 shared same multidrug resistance patterns, respectively. Notably, all isolates were resistant to AMP and 12 isolates (in addition to TJ1809, TJ3237 and TJ3277) were resistant to CHL, STR, SXT and SUL, which are commonly encoded by SXT/R391 ICEs. MD20140901 to MD20140905 were resistant to the first, second and third-generation cephalosporins; MD20140904 was also resistant to the fourth generation cephalosporin.  All isolates positive for resistant related genes (floR, STR, SUL and SXT) at ICEs, were phenotypic resistant to their drugs (chloramphenicol, streptomycin, sulfisoxazole and trimethoprim-sulfamethoxazole), by contrast, even if isoates were phenotypic resistant to those four drugs, their ICEs not always carried those resistant related genes (Table 1 and Fig. 2).

Transfer ability.
To test the transfer ability of the Proteus isolate ICEs, we selected five ICE-carrying isolates (08MAS2213, 08MAS1586, 09MAS2407, TJ1809, and TJ3335) for the mobility test. Transconjugants were obtained with a transfer frequency of 5.0 × 10 −6 (TJ3335) to 2.5 × 10 −2 (08MAS1586) per recipient cell ( Table 1). The transconjugants were confirmed by PCR detection of the int SXT and antibiotic resistance genes (strA/B, sul2, floR, and dfrA) in the recipient cells.

Discussion
In this study, the ICEs contained in Proteus isolates showed high diversity compared to those carried in a variety of other bacterial species. Our results indicated that SXT/R391 ICEs presented a strong ability to transmit among different bacterial species as a type of self-transmissible mobile genetic element. This study revealed the epidemiology of the spatio-temporal prevalence of ICEs in Proteus. The occurrence and dispersion of Proteus isolates from different regions in China revealed the occurrence and widespread distribution of ICEs among Proteus. Furthermore, the various ICEs conferred phenotypes such as multidrug and heavy metal resistance to their host strains. To date, at least 89 SXT/R391-family ICEs have been identified (http://db-mml.sjtu.edu.cn/ ICEberg/) 26 . Most ICEs have been investigated in V. cholerae 4,27-31 , which is the aetiological agent of the diarrhoeal disease cholera. The ICEs reported in Proteus strains including R997 (India) 32 , ICEPmiUSA1 (America) 33 , ICEPmiJpn1(Japan) 16 , ICEPmiSpn1(Spain) 24 and ICEPmiChn1(China) 17 . In this study, our fifteen ICEs revealed six distinct clusters and were positioned in different branches of the phylogenetic tree. The Blastn analysis in Genbank showed that closely related ICEs might be different. Additionally, five of our ICEs grouped into three distinctive clusters representing novel ICEs (Fig. 1), which increases the number of the SXT/R391 family members. Consequently, our results indicated that abundant genetic diversities and variable types of ICEs were ubiquitous in Proteus strains, compared with the ICEs in V. cholerae and other Enterobacteriaceae. The ICE types in the Proteus strains in our study were different from a recent investigation of the SXT/R391 ICEs in P. mirabilis isolates from food-producing animals in China 17 , which included two types of ICEs (ICEPmiJpn1 and ICEPmiChn1). We suggest that the use of a more extensive source of isolates (food and clinical samples from different cities) might contribute to the varieties of ICEs in Proteus isolates. Regardless of the diversity of the hosts and locations, the ICEs of the SXT/R391 family share a common structure and contain 52 conserved core genes that mediate integration, recombination, DNA repair and conjugative transfer 12 . In this study, the ICE-harbouring Proteus strains shared a common structure even though they were isolated from different cities in China and were at different evolutionary stages. However, the 15 ICEs were closely related to many different ICEs derived from Alteromonas macleodii, A. mediterranea, V. cholerae, V. alginolyticus, P. mirabilis and Providencia stuartii. They share much different query score and identity at nucleotide level (Table 1), consequently, these ICEs formed unique variable regions. Antibiotic resistance genes were typically present within VRIII in ICEs reported in previous studies. Our results showed that most of the multidrug resistance genes were present in this region, including the genes encoding resistance for sulfamethoxazole, trimethoprim, chloramphenicol, and streptomycin that were first described in the SXT of V. cholerae O139 MO10 7 . A class A β -lactamase gene was found, which has been reported in the plasmid from P. mirabilis (GenBank: KX228735), and the ICE from V. parahaemolyticus 34 . However, further research will be necessary to confirm the relationship of this gene with resistance to β -lactams. Five hotspots were located in s043-traL (HS1), traA-s054 (HS2), s073-traF (HS3), traN-s063 (HS4), and s026-traI (HS5). Variable genes were identified and predicted to encode restriction-modification systems, endonucleases, which may provide protection from invasion by foreign DNA 12 . ICEs without any antibiotic resistances were described 12 and determinants for antibiotic resistance were not found in ICEPmiCHN1809. However, whether the successful existence of the SXT/R391 ICEs is related exclusively to the appearance of resistance determinants is inconclusive. Other genes found in the variable regions, including genes encoding unknown functions, may be related to the enhancement of the adaptability of ICEs 12 .
In this study, the fifteen ICEs were grouped into six clusters by the phylogenetic analysis and therefore may be representative of six different evolutionary origins. Six ICEs were placed in cluster I, these ICEs were identical to ICEPmiChn1, which is a recently reported ICE contained in P. mirabilis that was isolated from a faecal sample from chicken in Hubei, China, in 2013 17 . ICEPmiChn1 was a predominant SXT/R391 family member in clinical Proteus sources in China and probably experienced horizontal transmission from food to humans. Because the six ICEs and the ICEPmiChn1-containing strains were isolated during and after 2013 (Supplementary Tables S1 and S2), these ICEs may have appeared during the latest period of their evolution and then presented an extensive distribution in China. Three ICEs (cluster II in Fig. 1) with references including ICEs from V. cholerae strain ICDC-4210 (isolated from Jiangxi, China, in 1999) 31 and P. mirabilis strain HI4320 (USA, 1986) 33 , indicated that the ICEs of this cluster may have appeared as early as 1986 and have been transmitted among different bacterial species and countries. Similarly, ICEPvuCHN2213 in cluster III contained in a P. vulgaris strain isolated from food was closely related to the ICE MED64 25 (from the Aegean Sea near Lebanon, 2000), which was between clusters I and II. Interestingly, five novel ICEs in this study were grouped into three distinctive clusters (IV, V and VI in Fig. 1) with no references in the three clusters, which might indicate that they were independently acquired by Proteus strains. Generally, the phylogenetic analysis suggested that ICEs could be transmitted among Proteus, V. cholerae and other Enterobacteriaceae and might share a common ancestor although they evolved independently. The acquired ICEs in Proteus were not species-specific under certain conditions; for example, to adapt to the environment and to facilitate survival under selective pressure, Proteus strains are able to acquire ICEs from different sources and evolutionary stages 35 .
The majority of known SXT/R391 ICEs contain four types of antibiotic genes (strA/B, sul2, floR and dfrA) that confer resistance to streptomycin, sulfamethoxazole, chloramphenicol and trimethoprim, respectively. R391 contains a kanamycin-encoding gene 11 and other new antibiotic resistance genes of known SXT/R391 ICEs carried The upper indicate backbone, represents the 52 conserved core genes of the SXT/R391 family ICEs. Below the common structure, indicate five hotspots (HS1-5) and two variable (III and IV) regions of the fourteen ICEs (except for ICEPmiCHN3277, which was incompletely assembled). Drug-resistance gene (red triangle); the restriction-modification system (RM) (yellow triangle); β -lactamase encoding gene (pink triangle); mercury resistance gene (orange triangle); other genes (blue triangle). Detailed annotated genes from all fifteen ICEs are listed in Supplementary Table S3.
Scientific RepoRts | 6:37372 | DOI: 10.1038/srep37372 have also been found (e.g. the cephalosporin resistance gene bla CMY-2 16 and rifampicin resistance gene) 13 . In this study, isolates positive for resistant related genes at SXT/R391 ICEs were not always consistent with their phenotypic resistance (Table 1), this result indicated the phenotypic resistance may not be associated with genes encoded by ICEs, determinants of other mobile genetic elements, like plasmid 36 , transposon 37 and genomic island 38 within Enterobacteriaceae, are also mediated drug resistance. However, the role of ICEs in the acquisiton and transmission of antibiotic resistance should not be neglected, our study have displayed high transfer ability of ICEs from Proteus isolates to recipient cells. In addition, the 15 SXT/R391 ICEs carried not only the four common resistance genes mentioned above, nine SXT/R391 ICEs also carried a class A β -lactamase gene. In contrast, no dominant β -lactamase genes are carried by ICEPmiChn1 17 , suggesting that the class A β -lactamase gene might have been obtained via horizontal gene transfer or recombination. In summary, our study reported multi-drug resistance, including the increasing prevalence of class A β -lactamase-producing P. mirabilis, which is consistent with the trends in Japan 16 , Taiwan 39 and other reports 24,40 . In conclusion, our results present abundant genetic diversity and multidrug resistance of ICEs carried by Proteus strains from both food sources and diarrhoeal patients. The SXT/R391 ICEs could be transferred between Proteus and other Enterobacteriaceae, thereby conferring resistance to the host and facilitating bacterial survival in the environment. Therefore, we need to strengthen the continuous monitoring of antimicrobial resistance and related mobile elements among Proteus isolates.