Identification of IncA/C Plasmid Replication
Steven J. Hancock , a , b Minh-Duy Phan , a , b Kate M. Peters , a , b Brian M. Forde , a , b Teik Min Chong , c Wai-Fong Yin , c Kok-Gan Chan , c David L. Paterson , d Timothy R. Walsh , e Scott A. Beatson , a , b Mark A. Schembria , b Australian Infectious Diseases Research Centre , University of Queensland , Brisbane , Australiaa ; School of Chemistry and Molecular Biosciences , University of Queensland , Brisbane , Australiab ; Faculty of Science , Division of Genetics and Molecular Biology , Institute of Biological Sciences , University of Malaya , Kuala Lumpur , Malaysiac ; University of Queensland Centre for Clinical Research , Brisbane , Australiad ; Department of Medical Microbiology and Infectious Disease , Cardiff University , Cardiff , United Kingdome 
ABSTRACT Plasmids of incompatibility group A/C ( IncA/C ) are becoming increasingly prevalent within pathogenic Enterobacteriaceae . 
They are associated with the dissemination of multiple clinically relevant resistance genes , including bla and CMY blaNDM . 
Current typing methods for IncA/C plasmids offer limited resolution . 
In this study , we present the complete sequence of a blaNDM-1-positive IncA/C plasmid , pMS6198A , isolated from a multidrug-resistant uropathogenic Escherichia coli strain . 
Hypersaturated transposon mutagenesis , coupled with transposon-directed insertion site sequencing ( TraDIS ) , was employed to identify conserved genetic elements required for replication and maintenance of pMS6198A . 
Our analysis of TraDIS data identified roles for the replicon , including repA , a toxin-antitoxin system ; two putative partitioning genes , parAB ; and a putative gene , 053 . 
Construction of mini-IncA/C plasmids and examination of their stability within E. coli confirmed that the region encompassing 053 contributes to the stable maintenance of IncA/C plasmids . 
Subsequently , the four major maintenance genes ( repA , parAB , and 053 ) were used to construct a new plasmid multilocus sequence typing ( PMLST ) scheme for IncA/C plasmids . 
Application of this scheme to a database of 82 IncA/C plasmids identified 11 unique sequence types ( STs ) , with two dominant STs . 
The majority of blaNDMpositive plasmids examined ( 15/17 ; 88 % ) fall into ST1 , suggesting acquisition and subsequent expansion of this blaNDM-containing plasmid lineage . 
The IncA/C PMLST scheme represents a standardized tool to identify , track , and analyze the dissemination of important IncA/C plasmid lineages , particularly in the context of epidemiological studies . 
KEYWORDS uropathogenic E. coli, IncA/C plasmid, functional genomics, New Delhi metallo-beta-lactamase, plasmid multilocus sequence typing
IncA/C plasmids are large , low-copy-number , broad-host-range plasmids with varying capacities for conjugation ( 1 ) . 
These plasmids represent an increasing threat to public health due to their association with the dissemination of the blaCMY cephalosporinase genes ( 2 ) and more recently the blaNDM metallo-beta-lactamase genes ( 3 -- 5 ) . 
The first IncA/C plasmids were isolated from aquatic host species , including the fish pathogen Aeromonas salmonicida ( 6 ) , and pandemic strains of Vibrio cholerae ( 7 , 8 ) . 
However , recently , there has been a significant increase in the isolation of IncA/C plasmids from Enterobacteriaceae , including Salmonella ( 9 ) , Klebsiella pneumoniae ( 10 ) , and Escherichia coli ( 11 ) . 
Plasmids from the IncA/C group were discovered more than 40 years ago and initially assigned to two separate groups , namely , IncA ( RA1 ) ( 6 ) and IncC ( 12 ) . 
Subsequent investigations into compatibility , exclusion , and phage sensitivity provided strong evidence for combining these two groups into a single group named IncA/C ( 12 -- 14 ) . 
Molecular analysis of the IncA/C replicon has again split this group into two distinct types , A/C1 ( RA1 ) and A/C2 ( 15 ) . 
The A/C2 type comprises the vast majority of IncA/C plasmids sequenced to date ( 1 ) . 
Plasmid backbone comparisons have informed the subtyping of IncA/C2 into type 1 and type 2 plasmids ( 16 ) . 
The two types differ in several ways , including two replacement regions ( R1 and R2 ) that lie within rhs and a large coding sequence ( CDS ) in transfer region 1 , respectively , as well as the presence or absence of two small segments ( i1 and i2 ) ( 16 ) . 
Common features are also found in relation to resistance gene content , for example , the vast majority of type 1 plasmids possess the antimicrobial resistance island A ( ARI-A ) located within rhs and an ISEcp1-blaCMY insertion within the large CDS in transfer region 1 ( 16 , 17 ) . 
An additional resistance island ( ARI-B ) , located upstream of the par locus , is found in both type 1 and 2 plasmids but is not always present . 
The IncA/C replicon was first defined and characterized using the archetype plasmid RA1 . 
Thirteen direct repeats ( iterons ) are located downstream of the repA replication gene , similar to IncP plasmids ( 18 ) . 
Both repA and the iterons are required for IncA/C replication ( 19 ) . 
Similar to other iteron-controlled replicons , there is an imperfect inverted repeat upstream of repA ( 18 ) , suggesting autoregulation ( 20 ) . 
IncA/C plasmids also possess a putative toxin-antitoxin ( TA ) system . 
The TA genes are strongly transcribed , suggesting the system is functionally active in the postsegregation killing of plasmid-free progeny ( 21 ) . 
Moreover , attempts to construct deletion mutants of the antitoxin component proved to be lethal to the cell ( 22 ) . 
Partitioning systems are one of the most important factors that contribute to the stable inheritance of large , low-copy-number plasmids ( 23 ) . 
Partitioning typically involves three components : a cis-acting DNA binding site ( centromere ; parS ) , a centro-mere binding protein ( ParB ) , and an NTPase ( ParA ) ( 24 ) . 
Together , they facilitate the correct positioning of plasmid molecules during cell division to increase plasmid retention ( 25 ) . 
Partitioning systems are classified into different groups based on the characteristics of the NTPase ( 24 ) . 
The IncA/C ParA protein contains a Walker-type ATPase , indicating IncA/C plasmids possess a type I partitioning system ( 1 , 25 ) . 
This ParA protein has similarity to ParA of IncP plasmids , while IncA/C ParB contains both ParB and KorB domains ( 1 ) . 
These genes are transcribed at low levels , similar to repA ( 21 ) . 
Another putative partitioning gene , stbA , is found in a separate genetic location . 
StbA has similarity to the ParM partitioning protein from the IncFII plasmid NR1 ( 1 ) . 
In addition to these elements , IncA/C plasmids carry a number of other genes putatively involved in replication , including kfrA and ter ( 1 ) . 
However , the functions of these genes have not been experimentally determined . 
Transposon-directed insertion site sequencing ( TraDIS ) , along with other , similar techniques , including Tn-seq ( 26 ) , INseq ( 27 ) , and HITS ( 28 ) , is a high-throughput whole-genome screening method used to perform bacterial functional genomic analyses ( 29 , 30 ) . 
A typical TraDIS experiment examines a highly saturated transposon mutant library under a condition of interest , with pre - and postselection libraries subjected to deep sequencing to simultaneously identify all of the transposon insertion sites . 
After selection , the lack of insertions within a gene is used to determine the importance of that gene for survival under the condition tested . 
The technique has been applied to identify genes that enable the maintenance and transmission of the IncI1 plasmid pESBL ( 31 ) and the essential genes of the IncF plasmid pEC958 ( 32 ) . 
Here , we employed TraDIS to identify genetic elements involved in the replication and maintenance of the IncA/C plasmid group . 
These experimentally validated elements provided a framework for development of a novel plasmid multilocus sequence typing ( PMLST ) scheme for tracking this important plasmid group . 
RESULTS
Genomic analysis of the carbapenem-resistant E. coli strain MS6198 . 
MS6198 is a carbapenem-resistant uropathogenic Escherichia coli ( UPEC ) strain . 
MS6198 is also nonsusceptible to multiple other antibiotics , including beta-lactams , nalidixic acid , ciprofloxacin , gentamicin , kanamycin , sulfamethoxazole , trimethoprim , tetracycline , and tobramycin ( see Data Set S1 in the supplemental material ) . 
The complete genome of MS6198 was determined and shown to consist of a circular chromosome comprising 5,176,750 base pairs ( 51.59 % G-C content ) . 
In silico typing assigned MS6198 to sequence type 648 ( ST648 ) , which has been associated with a number of disease outbreaks in Asia and Europe ( 33 -- 35 ) . 
MS6198 contained a number of UPEC virulence factors , including genes encoding type 1 fimbriae , Ag43 , capsule , and several iron acquisition systems . 
A list of known UPEC virulence genes found in MS6198 is shown in Data Set S2 in the supplemental material . 
In addition , MS6198 contained 4 circular plasmids , pMS6198A ( IncA/C type ; 137,565 bp ) , pMS6198B ( IncFII type ; 128,428 bp ) , pMS6198C ( untypeable ; 98,242 bp ) , and pMS6198D ( IncI1 type ; 50,899 bp ) . 
Methylome analysis of MS6198 identified three distinct DNA recognition motifs , indicating the presence of at least three active adenine methyltransferase enzymes ( see Data Set S3 in the supplemental material ) . 
Characterization of an IncA/C multidrug resistance plasmid harboring the blaNDM-1 carbapenemase gene . 
We focused our study on characterization of the blaNDM-1-positive IncA/C plasmid pMS6198A . 
Plasmid pMS6198A belongs to type 1 of the IncA/C2 group and contains 172 CDSs classified into seven functional groups ( Fig. 1 ) . 
The plasmid contains the typical IncA/C replicon and additional putative maintenance genes , including parAB , parM , kfrA , and a putative TA system . 
Similarity searches of public sequence databases indicated that pMS6198A contains 19 conjugation genes , including the master regulators acaDC ( 22 ) . 
In addition to a typical ISEcp1-blaCMY-6 insertion within transfer region 1 , five resistance genes are located within ARI-A -- aacA4 , rmtC , blaNDM-1 , bleMBL , and sul1 -- along with a truncated qacEΔ1 gene . 
Plasmid pMS6198A contains 92 CDSs that have no assigned annotation . 
Sequence comparison with other IncA/C plasmids showed that pMS6198A displayed 99 % conservation over its entire sequence with two other blaNDM-1-positive IncA/C plasmids : pKP1-NDM1 ( KF992018 ) from Australia and pNDM10469 ( JN861072 ) from 
Canada . 
The genetic structure of blaNDM-1 in pMS6198A is highly similar to those of eight other IncA/C plasmids ( see Fig . 
S1 in the supplemental material ) . 
A full complement of transfer genes , as defined previously ( 1 ) , were present in pMS6198A . 
Accordingly , pMS6198A was transferable by conjugation in static liquid culture at 37 °C to the recipient strain E. coli J53 at a frequency of 10 3 transconjugants per donor . 
Transfer of pMS6198A to J53 was confirmed by PCR amplification of the IncA/C replicon . 
Antibiotic resistance profiling of the transconjugant strain ( MS6614 [ see Data Set S1 in the supplemental material ] ) showed that pMS6198A was capable of conferring resistance to multiple antibiotics , including cefotaxime-clavulanic acid , ceftazidime-clavulanic acid , piperacillin-tazobactam , amoxicillin-clavulanic acid , cefoxitin , cefpodoxime , ceftriaxone , cephalothin , ampicillin , gentamicin , kanamycin , imipenem , meropenem , ertap-enem , and tobramycin . 
Purification of pMS6198A from MS6614 and electrotransformation into E. coli TOP10 were performed , with subsequent analysis of the transformed TOP10 strain by PCR , conjugation , and antibiotic resistance profiling indicating the integrity of the plasmid was maintained ( see Data Set S1 in the supplemental material ) . 
Thus , all subsequent analysis of pMS6198A was performed using plasmids purified from MS6614 . 
Identification of genes required for pMS6198A maintenance and replication . 
To identify the genes required for the maintenance and replication of pMS6198A in E. coli , we employed in vitro mutagenesis in combination with TraDIS as shown in Fig. 2 . 
First , in vitro mini-Tn5 mutagenesis of pMS6198A was carried out to create a highly saturated mutant plasmid DNA library . 
This library was transformed into E. coli TOP10 by electroporation and subsequently grown in the presence of chloramphenicol to select for pMS6198A : : mini-Tn5-containing transformants . 
This process was performed in duplicate , resulting in two saturated libraries , each of which contained approximately 10,000 transformants . 
Purified plasmid DNA was extracted from both libraries and analyzed by TraDIS . 
Examination by inverse PCR of a subset of individual Tn5 mutant colonies was performed to investigate the randomness of insertion and the number of insertions per plasmid molecule . 
The majority of mutants examined ( 17/18 ; 94 % ) contained a single insertion , each at a different location on the plasmid ( see Fig . 
S2 in the supplemental material ) . 
TraDIS identified a total of 10,178 unique insertion sites in pMS6198A from the two libraries , which was equivalent to an average of one insertion site every 13.52 bp . 
Correlation tests of insertion sites per gene for each library showed high reproducibility for the TraDIS sequencing method ( R2 0.99 ) ( see Fig . 
S3 in the supplemental material ) . 
The relative abundance of reads mapped to mini-Tn5 insertion sites within each gene ( expressed in reads per kilobase per million [ RPKM ] ) was calculated ( Fig. 3A ; see Data Set S4 in the supplemental material ) . 
This enabled us to identify genes required for the maintenance and replication of pMS6198A , as plasmids containing mutations in such genes would be lost and thus underrepresented in the libraries , reflected as a low RPKM value . 
The repA gene is known to be required for IncA/C plasmid replication ( 19 ) , and we observed a relative abundance of mini-Tn5 insertions in repA ( 418 RPKM ) ( Fig. 3Ci ) significantly lower than the average insertion abundance for all pMS6198A genes ( 7,222 RPKM ) . 
Therefore , we used 418 RPKM as a biological threshold for genes required for replication and maintenance of pMS6198A . 
Six additional genes were identified with insertion abundances lower than this threshold : 022 , parA , parB , 053 , tnpA , and rmtC ( Fig. 3B ) . 
The putative TA system of IncA/C plasmids belongs to the tad-ata-like family found in a range of different genetic elements , including ICE SXT , Enterobacteria phage N15 , a genomic island of E. coli EDL933 , and the plasmid pAMI2 ( 36 ) . 
The 022 gene ( here referred to as ata , for antitoxin of addiction system ) lies immediately adjacent to the 023 gene ( referred to as tad , for toxin of addiction system ) ( Fig. 3Cii ) . 
Both tad and ata are 100 % conserved in all fully sequenced IncA/C2 plasmids but absent from IncA/C1 plasmids . 
Tad is a member of the Gp49 family proteins ( Pfam identifier PF05973 ) , which is comprised of known toxins . 
In the phage N15 , Gp49 is controlled by the adjacently encoded Gp48 protein ( UniProt accession no . 
O64356 ) , which contains a helix-turn-helix ( HTH ) DNA binding domain ( 36 ) . 
Analysis of the Ata amino acid sequence showed it is 24 % identical and 50 % similar to that of Gp48 ( 71 % coverage ) and possesses an HTH domain ( Pfam identifier PF13744 ) , suggestive of a DNA binding function . 
Our TraDIS data showed that mutation of ata is strongly underrepresented , while mutation of tad is tolerated . 
Taken together , our data suggest that ata and tad encode an active TA system , in which Ata is the antitoxin . 
Three adjacent genes , parA ( 051 ) , parB ( 052 ) , and 053 , were also identified in our TraDIS analysis ( Fig. 3Ciii ) . 
The three genes are conserved in all fully sequenced IncA/C plasmids . 
ParA belongs to an ATPase family ( Pfam identifier PF13614 ) , while ParB contains both ParB ( Pfam identifier PF02195 ) and KorB ( Pfam identifier PF08535 ) domains . 
Based on bioinformatics analysis , parA and parB are homologous to IncP partitioning genes . 
The third gene in the locus ( 053 ) encodes a hypothetical protein of 90 amino acids ( aa ) containing a winged HTH domain ( Pfam identifier PF09904 ) , suggestive of a DNA binding function . 
Additionally , 053 is present at the same genetic location ( after parB ) in all IncA/C plasmids . 
Two other genes , tnpA and rmtC , were identified in our TraDIS screen ( Fig. 3Civ and v ) . 
The tnpA gene encodes a transposase from the insertion sequence ISEcp1 ( 37 ) , while rmtC encodes a 16S ribosomal methyltransferase ( UniProt entry Q33DX5 ) that confers resistance to many aminoglycoside antibiotics . 
These two genes occupy regions of pMS6198A with the lowest G-C content ( see Fig . 
S4 in the supplemental material ) that may be associated with an insertion bias ( 38 ) . 
Furthermore , the genes are not conserved among completely sequenced IncA/C plasmids , being present in only 50 % and 11 % , respectively . 
Based on these data , they are unlikely to be involved in the maintenance and replication of pMS6198A . 
The par locus of IncA/C plasmids , of which 053 is a crucial component , contributes to stability . 
TraDIS analysis identified two putative partitioning genes ( parA and parB ) and a region containing an open reading frame ( ORF ) immediately downstream of parB ( 053 ) as required for plasmid maintenance . 
Bioinformatics analysis revealed that the structural organization of parA-parB-053 is completely conserved in all sequenced IncA/C plasmids examined in this study . 
This led us to hypothesize that the partitioning system in IncA/C is comprised of three components that are all required for plasmid stability in E. coli . 
To provide evidence to support our hypothesis , we constructed pMS6198A-derived mini-A/C plasmids containing different versions of the partitioning locus and examined their stability in E. coli strain MG1655 . 
The mini-A/C plasmid contained the IncA/C replicon , as previously described ( 1 , 18 , 19 ) . 
A selectable marker ( cat cassette , conferring chloramphenicol resistance ) was included , along with two transcriptional terminators to prevent transcriptional read-through from the cassette . 
Two variations of the partitioning locus were incorporated into this mini-A/C plasmid , generating pMAC2 ( parAB ) and pMAC3 ( parAB plus 053 ) ( Fig. 4A ) . 
The stability of pMS6198A was initially examined by growth in the absence of selection for three serial passages , each incorporating 10 generations . 
After 30 generations , no plasmid loss was observable ( see Fig . 
S5 in the supplemental material ) , highlighting the high stability of the parent plasmid . 
The same experiment was used to assess the stability of pMAC2 and pMAC3 . 
This showed that pMAC2 was less stable than pMAC3 : even at time zero , only 50 % of cells retained pMAC2 , and by 30 generations , this had dropped to 2 % ( see Fig . 
S5 in the supplemental material ) . 
In contrast , pMAC3 started at 100 % and was reduced to 64 % after 30 generations ( see Fig . 
S5 in the supplemental material ) . 
The difference in plasmid stabilities observed between the starting populations of pMAC2 and pMAC3 ( time zero ) was addressed by mixing cells harboring pMAC3 in a 1:1 ratio with plasmid-free cells , thus mimicking the starting population of cells harboring pMAC2 . 
Analyses using this equivalent starting population resulted in plasmid stability profiles very similar to those observed in the single-strain experiment , with pMAC2-harboring cells reduced to 5 % after 30 generations compared to significantly higher ( 57 % ) retention of pMAC3 ( Fig. 4B ) . 
The TraDIS-identified maintenance genes are conserved in all IncA/C plasmids . 
Plasmid multilocus sequence typing was originally developed for the typing of large collections of plasmids using a set of conserved genes ( 39 ) . 
In the context of IncA/C plasmid typing , our TraDIS data have provided a defined subset of genes with essential plasmid maintenance functions that could be used in the development of a PMLST scheme . 
To provide a framework for this analysis , a collection of 82 complete IncA/C plasmid sequences available in the GenBank database were examined for overall sequence conservation ( see Data Set S5 in the supplemental material ) . 
A total of 28 genes were completely conserved within this collection , which included the four genes identified by TraDIS ( repA , parA , parB , and 053 ) ( Table 1 ) . 
The concatenated sequences of all 28 fully conserved genes from each plasmid were used to build a maximumlikelihood tree ( Fig. 5 , left ) . 
Interestingly , this analysis identified multiple previously defined hybrid plasmid groups , including pYR1 and p39R861-4 , and also differentiated between type 1 and 2 A/C2 plasmids ( 1 , 16 , 40 ) , with type 1 plasmids also separated into three distinct groups . 
Using this as a baseline , the discriminatory power of repA , parA , parB , and 053 was examined by using the concatenated sequences of the four genes from each plasmid to build a maximum-likelihood tree ( Fig. 5 , right ) . 
The overall topology of the tree is similar to our analysis of IncA/C plasmids using all 28 conserved genes , including a split between IncA/C1 ( group 5 ) and IncA/C2 ( groups 1 to 4 ) . 
The IncA/C2 lineage was further split into four distinct groups ; groups 1 and 2 comprised type 1 plasmids , and group 3 included both type 1 and type 2 plasmids , while group 
4 represented the hybrid plasmid pYR1 . 
Although the resolution of the 28 conserved genes separated hybrid plasmid p39R861-4 and both type 1 and type 2 plasmids , their similarity in the sequences of repA-parA-parB-053 clustered the plasmids together in group 3 . 
Furthermore , the preservation of groups 1 , 2 , 4 , and 5 between the two analyses highlights them as distinct plasmid lineages . 
Taken together , the analysis demonstrates that the sequences of the four essential genes identified by TraDIS are sufficient to capture the phylogenetic relatedness of IncA/C plasmids and could be used for molecular typing . 
Development of an IncA/C PMLST scheme . 
As the repA , parA , parB , and 053 genes could distinguish different groups of IncA/C plasmids , we used these biologically 
Primer Sequence ( 5 = -- 3 =) Primer information size ( bp ) a repA-F AAGAGAACCAAAGACAAAGAC Amplify repA 982 repA-R GCTGCTTACGCTTGTTGGA parA-F AAAAGTAATCAGCTTCGCCA Amplify parA 780 parA-R TAGCCCACCTTCTCTAATAG parB-F TGTCCGAACTTGCTAAAGC Amplify parB 1,128 parB-R CTGACACAGGCACATGAA 053-F AGATCTCACAGGACATGAA Amplify 053 250 053 - R TTCAAGAACGAAGACCTGT repA-Seq1 TGGAGTTCGTACAGAGTGA Sequence 5 = region of repA fragment NA repA-Seq2 GCTCCAGCTTCTTCCCGAT Sequence 3 = region of repA fragment NA parB-Seq1 CACACAGTCAGGTAGCTT Sequence 5 = region of parB fragment NA parB-Seq2 AAGCTACCTGACTGTGTG Sequence central region of parB fragment NA parB-Seq3 GATGCTCTTCCTCCTCTG Sequence 3 = region of parB fragment NA validated genes to develop a PMLST scheme for IncA/C plasmids . 
Amplification and sequencing primers were designed to target conserved regions within each essential gene ( Table 2 ) . 
PCR and Sanger sequencing using these primers were performed on pMS6198A to validate the methodology . 
Using the sequences obtained in silico from 82 IncA/C plasmids , different alleles for repA ( n 5 ) , parA ( n 6 ) , parB ( n 7 ) , and 053 ( n 3 ) were identified , which together form 11 STs , as shown in Fig. 6 ( see Data Set S6 in the supplemental material ) . 
The minimum spanning tree comprises two singletons , ST10 ( pYR1 ; IncA/C2 ) and ST11 ( RA1 ; IncA/C1 ) , with the remaining nine STs linked together as single-locus variants ( SLVs ) or double-locus variants ( DLVs ) . 
ST3 is the largest group and includes 53 plasmids . 
ST3 connects to ST4 , ST5 , ST6 , ST7 , ST8 , and ST9 as SLVs , forming a clonal complex with ST3 as the founder ( ST3 clonal complex [ ST3CC ] ) . 
ST3 links with ST2 as a DLV , and ST2 connects to ST1 as an SLV . 
ST1 is the second largest group and comprises 20 plasmids . 
ST1 and ST3 together account for 89 % of the total IncA/C plasmids investigated . 
To provide higher-resolution typing applicable to next-generation sequencing data , a core gene PMLST ( cgPMLST ) scheme was also constructed by extending the 4-gene PMLST to 28 conserved genes ( see Data Set S7 in the supplemental material ) . 
The four loci shared between the two schemes allow backward compatibility from cgPMLST to PMLST . 
Thus , cgPMLST is capable of subtyping 11 STs into 35 subgroups , including 4 subgroups for ST1 ( ST1 .1 to ST1 .4 ) , 22 subgroups for ST3 ( ST3 .1 to ST3 .22 ) , and 1 subgroup for each of the remaining STs ( see Data Set S7 and Fig . 
S6 in the supplemental material ) . 
IncA/C PMLST highlights a lineage of blaNDM-harboring plasmids . 
To determine if the distribution of antibiotic resistance genes among IncA/C plasmids is associated with STs , the resistance gene content of each plasmid was overlaid with the PMLST phylogenetic scheme ( Fig. 7 ) . 
Examination of the resistance gene profiles highlighted distinct patterns between ST1 and ST3CC . 
ST1 predominantly harbors blaCMY-6 , while 
ST3CC plasmids mainly possess the blaCMY-2 variant ( Fig. 7 ) . 
ST3CC also exhibits a higher prevalence of tetracycline and streptomycin resistance genes . 
ST1 is strongly associated with the carriage of blaNDM ( 15/20 plasmids ; 75 % ) , while only two plasmids outside ST1 carry blaNDM ( one each from ST3 and ST6 ) . 
Comparison of the blaNDM genetic location has shown multiple genetic organizations ( 41 -- 46 ) . 
Within our database , we observed a total of seven distinct blaNDM structures ( see Fig . 
S1 in the supplemental material ) . 
Six of them are found in individual plasmids ( pNDM-SAL , ST1 ; pNDM-1_Dok01 , ST1 ; pEC2-NDM-3 , ST1 ; pNDM15-1078 , ST1 ; pRH-1238 , ST3 ; and pMR0211 , ST6 ) , and one has been observed within ARI-A of nine plasmids ( 45 , 47 ) . 
All nine plasmids belong to ST1 , subgroups ST1 .2 and ST1 .4 ( see Fig . 
S6 in the supplemental material ) . 
Overall , non-ST1 blaNDM structures are distinct from all others , while the presence of similar structures in 9/15 ST1 plasmids suggests the successful dissemination and diversification of the plasmid lineage . 
DISCUSSION
Carbapenem-resistant Enterobacteriaceae have been recognized as an urgent threat to human health ( 48 ) . 
Infections caused by these bacteria are often resistant to almost all clinically available antibiotics and are frequently associated with poor health outcomes . 
New Delhi metallo - - lactamase is a recently emerged carbapenemase first described in 2009 ( 49 ) . 
Since then , several IncA/C plasmids carrying the blaNDM-1 gene ( 4 , 41 , 47 , 50 -- 52 ) , and recently blaNDM-3 ( 45 ) , have been reported . 
Plasmids of incompatibility group A/C have been known for more than 40 years , but they have only recently gained increased interest due in part to their emergence as the major plasmid type carrying the cephalosporinase gene blaCMY ( 2 ) . 
However , the broad-host range characteristic of IncA/C plasmids and their roles in the dissemination of multiple antibiotic resistance genes , including blaCMY and blaNDM , are underappreciated . 
Here , we have used a validated set of essential genes to develop a high-resolution typing scheme to monitor the spread and transmission of IncA/C plasmids . 
Previous studies on the replicon of RA1 , the prototypical IncA/C plasmid , showed that IncA/C plasmid replication is mediated by the autoregulated repA gene and an iteron upstream of a DnaA box ( 19 ) . 
Indeed , our mutagenesis analysis confirmed a requirement for repA and its adjacent iteron region in pMS6198A ( Fig. 3Ci ) . 
This validated our method to identify required genetic components and also served as a reference point for identification of six other genes , namely , ata , parA , parB , 053 , tnpA , and rmtC . 
The addiction system of IncA/C plasmids has been the subject of various studies , but its contribution to IncA/C plasmid stability has not been fully established . 
Recent transcriptome analysis of pAR060302 showed that the system is strongly transcribed ( 21 ) . 
Furthermore , multiple attempts to mutate ata were ultimately unsuccessful ( 22 ) . 
Our TraDIS data confirmed that ata mutants were highly underrepresented . 
This is likely a result of uncontrolled toxin expression leading to cell death and demonstrates that the system is active in pMS6198A . 
The activity of the system may contribute to the variation in stability between the parent plasmid , pMS6198A , and pMAC3 ( see Fig . 
S5 in the supplemental material ) . 
Moreover , it is possible that additional stability elements may be present on pMS6198A that exert subtle effects undetected by our stringent TraDIS threshold . 
A putative partitioning locus is present in all IncA/C plasmids . 
It has been noted that the parA and parB genes share similarity to IncP plasmid partitioning and regulatory elements ( 1 ) . 
Our TraDIS data strongly support a role for these genes in the maintenance and stability of IncA/C plasmids . 
Additionally , we identified the adjacent ORF , 053 , as a novel element of the locus that has no IncP homolog . 
We showed that parAB alone were not sufficient for maintenance ( Fig. 4 , pMAC2 ) and that plasmid stability improved markedly with the addition of a 420-bp fragment containing ORF 053 ( Fig. 4 , pMAC3 ) . 
Our results invoke at least two hypotheses : that a parS cis-acting centromerelike site of the ParAB partitioning system is present within this region or that 053 encodes a novel partitioning protein . 
Further work is required to elucidate the mechanisms by which this 420-bp region contributes to plasmid stability . 
Plasmid pMS6198A carries a number of other genes with putative replication and maintenance functions , including kfrA ( 161 ) and stbA ( 174 ) . 
However , they were not identified by TraDIS analysis ( see Data Set S4 in the supplemental material ) , suggesting they are not critical to pMS6198A maintenance in E. coli . 
Conversely , tnpA and rmtC have functions unrelated to plasmid stability yet were identified . 
Both genes have low GC contents ( 34 % and 41 % , respectively , compared to an average of 52 % ) across the entire pMS6198A sequence . 
This is consistent with an overall increased mini-Tn5 insertion frequency in high - versus low-GC regions of pMS6198A ( see Fig . 
S4 in the supplemental material ) . 
Thus , the identification of these genes may be the result of insertion bias of the Tn5 transposon in the in vitro mutagenesis reaction ( 38 ) . 
Deciphering phylogenetic relationships among plasmids can be challenging due to their mosaic nature . 
However , within one plasmid incompatibility group , it is expected that replication and maintenance machineries required for its biology should be conserved . 
This notion has been successfully applied to select genes for multilocus sequence typing schemes of plasmids from many Inc groups , including IncI1 ( 39 ) , IncHI1 ( 53 ) , IncHI2 ( 54 ) , and IncN ( 55 ) . 
Here , we propose a new PMLST scheme for IncA/C plasmids based on an experimentally validated set of essential genes . 
The loci selected for our PMLST scheme were based on the required genes identified by TraDIS analysis and supported by their presence in all IncA/C plasmids investigated . 
The scheme identified 11 sequence types , demonstrating higher resolution than current typing methods . 
We suggest that our PMLST scheme could be used to monitor dissemination and diversification patterns of IncA/C plasmids . 
Of the 11 IncA/C sequence types , the majority of plasmids belong to two main types : ST1 and ST3 . 
ST3 is the largest group , comprising 53 plasmids that form a clonal complex ( ST3CC ) with plasmids from ST4 , ST5 , ST6 , ST7 , ST8 , and ST9 ( Fig. 6 ) . 
The ST3CC plasmids were isolated from 1969 to 2015 in 19 countries and 16 species ( Fig. 6 ; see Data Set S5 and Fig . 
S7 in the supplemental material ) , highlighting their wide geographical distribution and broad host range characteristics . 
Interestingly , the IncA/C2 type 1 plasmids within ST3CC showed strong association with blaCMY ( 23/31 ; 74 % ) . 
As highlighted previously , the high frequency of blaCMY-positive IncA/C plasmids isolated from Salmonella enterica in the United States is an indication of sampling bias within the data set ( 1 ) . 
ST1 is the second-largest group in our data set , with 20 plasmids . 
The most striking feature of ST1 is the strong association with the carriage of blaNDM , with the majority of ST1 plasmids isolated from E. coli and K. pneumoniae . 
The clinical relevance of these species and carbapenem resistance has likely contributed to a sampling bias of ST1 plasmids . 
Nevertheless , it is tempting to hypothesize that ST1 is a newly emerged lineage of IncA/C plasmids , with carbapenem resistance enhancing its selection and dissemination . 
Further work is needed to test this hypothesis ; however , the presence of a common blaNDM structure in specific ST1 plasmid subgroups is supportive of the tenet . 
The observation of most concern here is the identification of ST1 in nine countries spanning four continents ( see Data Set S5 and Fig . 
S7B in the supplemental material ) , highlighting the urgent need for surveillance and control of this extensively drug-resistant plasmid lineage . 
More in-depth investigations within each ST may provide a better framework for analyzing the evolution of IncA/C plasmids . 
Inspired by the development of core genome MLST schemes ( 56 ) , we have also constructed a cgPMLST scheme for IncA/C plasmids using the 28 conserved genes identified in our database ( see Fig . 
S6 and Data Set S7 in the supplemental material ) . 
IncA/C cgPMLST is intended as a subtyping scheme complementary to PMLST to increase the discriminatory power suitable for plasmid epidemiology studies . 
The discriminatory power , measured by Hunter 's index ( 57 ) , of our IncA/C PMLST is 0.53 and is increased to 0.90 for cgPMLST . 
Other methods , such as that proposed by Harmer and Hall ( 58 ) , could also be used in conjunction with PMLST to subtype IncA/C plasmids . 
While sequence data from next-generation sequencing platforms , especially those from Pacific Biosciences ( PacBio ) , would provide full plasmid backbone sequences for in-depth epidemiology and high-resolution phylogeny analysis , such technologies remain out of reach for many laboratories in developing countries , where surveillance and control measures are most needed . 
Our PMLST scheme provides a robust method based on PCR and Sanger sequencing for the identification of major lineages of IncA/C plasmids . 
These major lineages can then be subtyped using cgPMLST when wholeplasmid sequence data are available . 
Like other MLST-based schemes , both of our schemes are compatible with freely available tools , such as SRST2 ( 59 ) and pMLST ( 60 ) , for in silico determination of plasmid STs , enabling the use of typing data across different settings . 
Our PMLST and cgPMLST schemes are also available on the public databases for molecular typing and microbial genome diversity ( PubMLST ) ( 79 ) . 
MATERIALS AND METHODS
Bacterial strains and growth conditions . 
E. coli MS6198 was isolated from the urine of a patient with a urinary tract infection in Haryana , India , in 2010 ( 61 ) . 
The E. coli strain J53 was provided by G. Jacoby ( 62 ) . 
The strains were routinely cultured at 37 °C under orbital shaking ( 250 rpm ) , in liquid or on solid lysogeny broth ( LB ) medium , supplemented with appropriate antibiotics . 
The following concentrations were typically used : ampicillin , 100 g/ml ; sodium azide , 100 g/ml ; meropenem , 1 g/ml ; and chloramphenicol , 30 g/ml . 
Electrocompetent cells were prepared , and transformations were performed as described previously ( 29 ) . 
All the strains were stored in 15 % glycerol at 80 °C . 
DNA purification and analysis . 
The PureLink HiPure Midiprep plasmid DNA purification kit ( Invitrogen ) was used to purify pMS6198A , while vector plasmids ( 10 kb ) were purified with the QIAprep Spin Miniprep kit ( Qiagen ) . 
Genomic DNA was obtained with MoBio 's Ultraclean microbial DNA isolation kit . 
DNA concentrations were quantified using a NanoDrop 2000 ( Thermo Scientific ) and/or Qubit 2.0 ( Life Technologies ) fluorometer . 
PCR and sequencing . 
The presence of plasmids was determined by PCR - based replicon typing ( 63 , 64 ) ; blaNDM was identified with primers 5 = - GGTTTGGCGATCTGGTTTTC-3 = and 5 = - CGGAATGGCTCATCACG ATC-3 = as previously described ( 65 ) , using One Taq polymerase ( New England BioLabs ) . 
All restriction enzymes , T4 ligase , and Antarctic phosphatase were purchased from New England BioLabs . 
All capillary sequencing reactions were prepared using BigDye Terminator mix and sequenced by the Australian Equine Genetics Research Centre ( AEGRC ) . 
A full list of primers used in this study is shown in Data Set S9 in the supplemental material . 
PMLST PCR and sequencing protocol . 
Amplification of the four loci used in PMLST was performed with Kapa HiFi DNA polymerase ( Kapa Biosystems ) using primers listed in Table 2 with the following cycling program : 95 °C for 3 min ; 25 cycles of 98 °C for 20 s , 60 °C for 15 s , and 72 °C for 30 s ; and a final extension of 72 °C for 3 min . 
Each amplicon was then purified using the QIAgen PCR purification kit and sequenced using BigDye Terminator v3 .1 cycle sequencing ( Life Technology ) with the appropriate sequencing primers listed in Table 2 . 
Disc diffusion and mating assays . 
The disc diffusion assay was performed and interpreted according to the Clinical and Laboratory Standards Institute guidelines ( 2014 ) . 
Antimicrobial discs were obtained from Becton Dickinson . 
For mating assays , the sodium azide-resistant E. coli strain J53 ( 62 ) was used as the recipient in all mating assays . 
Donor and recipient strains were grown to an optical density at 600 nm ( OD600 ) equal to 2.0 . 
The cells were then mixed at a ratio of 1:2 ( donors to recipients ) in LB and incubated at 37 °C for 2 h under static conditions . 
Total CFU of donors , recipients , and transconjugants were enumerated on LB agar plates with appropriate antibiotic selection ( ampicillin for donors and sodium azide for recipients ) , and the conjugation frequency was calculated as the number of transconjugants per donor . 
Plasmid stability assays . 
Time course stability assays were performed essentially as previously described ( 66 ) . 
Population counts were achieved by 10-fold serial dilutions and 5 - l drop test on LB agar with or without selection . 
For solid-medium stability assays , strains were grown overnight on LB agar supplemented with antibiotics , and then single colonies were suspended in 0.9 % NaCl . 
Population counts were achieved by 10-fold serial dilutions and 5 - l drop tests on LB agar with or without selection . 
In vitro transposon mutagenesis . 
A custom mini-Tn5 transposon containing a chloramphenicol resistance cassette ( Tn5-Cm ) was generated as previously described ( 30 ) . 
An in vitro plasmid mutant library was created by incubating 200 ng of pMS6198A DNA and equimolar Tn5-Cm DNA ( 1.588 ng ) with 1 l of Tn5 transposase ( 1 U / l ) from an EZ-Tn5 R6K ori/KAN -2 insertion kit ( Epicentre ) at 37 °C for 2 h . 
The reaction was stopped by adding 1 l 1 % SDS and heating at 70 °C for 10 min . 
The mutant plasmid library ( 1 l ) was transformed into 50 l of E. coli TOP10 electrocompetent cells . 
Cells carrying mutant plasmids were selected by plating on LB agar supplemented with chloramphenicol . 
Mutants were pooled by scraping colonies off agar plates into LB . 
After addition of glycerol to a final concentration of 15 % , the mutant library was stored at 80 °C . 
Inverse-PCR method . 
Purified DNA from individual mutants was digested with BanII for 2 h at 37 °C and heat inactivated at 65 °C for 20 min . 
This mixture was ligated with T4 DNA ligase overnight at 16 °C . 
This ligation mixture was used as the template for PCR using OneTaq and cat-specific primers 3748 and 3950 , with thermocycling : 94 °C for 1 min ; 30 cycles of 94 °C for 30 s , 62 °C for 30 s , and 68 °C for 3.5 min ; and a final extension time of 5 min . 
Transposon-directed insertion site sequencing . 
Illumina library preparation was performed using a Nextera DNA Sample Prep kit ( Illumina ) following the manufacturer 's instructions with modifications for TraDIS . 
Briefly , genomic DNA was fragmented and tagged with adapter sequence via one enzymatic reaction ( tagmentation ) . 
Following tagmentation , the DNA was purified using the Zymo DNA Clean and Concentrator kit ( Zymo Research ) . 
The PCR enrichment step was run using index primer 1 ( one index per sample ) and a custom transposon-specific primer , 4844 ( 5 = - AATGATACGGCGACCACCGAGATCTACACTA GATCGCaacttcggaataggaactaagg-3 = [ transposon-specific sequence is in lowercase ] ) to enrich for transposon insertion sites and allow multiplexing sequencing ; the thermocycler program was 72 °C for 3 min and 98 °C for 30 s , followed by 22 cycles of 98 °C for 10 s , 63 °C for 30 s , and 72 °C for 1 min . 
Each library was purified using Agencourt Ampure XP magnetic beads . 
Verification and quantification of the resulting libraries were calculated using a Qubit 2.0 fluorometer , a 2100 Bioanalyzer ( Agilent Technologies ) , and quantitative PCR ( qPCR ) ( Kapa Biosciences ) . 
All libraries were pooled in equimolar amounts to a final concentration of 3.2 nM and submitted for sequencing on the MiSeq platform at the Queensland Centre for Medical Genomics ( Institute for Molecular Bioscience , University of Queensland ) . 
The MiSeq sequencer was loaded with 12 pM of pooled library with 5 % PhiX spike-in and sequenced ( single-end ; 101 cycles ) using a mixture of standard Illumina sequencing primer and Tn5-specific sequencing primer 4845 ( 5 = - actaaggaggatattcatatggaccatggctaattcccatgtcagatgtg-3 =) . 
All sequence data analysis and insertion site mapping were performed as previously described ( 30 , 32 ) . 
The threshold for plasmid maintenance was set to the value of the repA gene ( 418 RPKM ) , in accordance with previous work that has demonstrated the gene is required for IncA/C plasmid replication ( 19 ) . 
Construction of mini-A/C plasmids . 
The replicon fragment was amplified with primers 7139/7140 from purified pMS6198A , the transcriptional terminator ( TT ) fragment gene was amplified with primers 7141/7142 from pQE30 , and the cat gene was amplified with primers 7143/7144 from pKD3 . 
The primers were designed so that adjacent fragments had 20-bp complementary overhangs , and a multiple-cloning site ( MCS ) was included between the replicon and TT fragments . 
All three fragments were mixed in equimolar ratios in a PCR using Kapa HiFi with thermocycling : 95 °C ; 30 cycles of 98 °C for 20 s , 66 °C for 20 s , and 72 °C for 6 min ; and a final extension of 72 °C for 8 min . 
The product of this reaction was electroporated into E. coli TOP10 . 
The desired product was confirmed by PCR screening and NheI digestion of purified plasmid DNA . 
The TT fragment contained two terminators , lambda t0 and rrnB t1 , separated by a cat gene . 
The cat gene was removed by PCR amplification of the plasmid using primers 7145/7146 , followed by NheI digestion 37 °C for 2 h and overnight ligation with T4 DNA ligase . 
The product was electroporated into E. coli TOP10 , and subsequent purification , PCR screening , and NheI digestion confirmed the desired product , which was referred to as pMAC1 . 
The partitioning fragments were amplified using Kapa HiFi with primers 7147/7149 for pMAC2 and 7147/7148 for pMAC3 . 
Each fragment was cloned into the MCS of pMAC1 using BamHI and HindIII restriction sites . 
The resultant plasmids ( pMAC2 and pMAC3 ) were confirmed to be correct by sequencing . 
Genome sequencing , assembly , and methylome analysis . 
Genomic DNA from MS6198 was sequenced on the PacBio RSII ( University of Malaya ) using the P4 polymerase and C2 sequencing chemistry . 
The raw sequencing data were assembled de novo using the hierarchical genome assembly process ( HGAP ) version 2 from the SMRT Analysis software suite ( version 2.3.0 ; Pacific Biosciences ) with default parameters . 
The assembled contigs were visually screened for overlapping sequences on their 5 = and 3 = ends using contiguity ( 67 ) . 
These overlapping ends were manually trimmed based on sequence similarity , and the contigs were circularized . 
The circularized contigs ( chromosome and plasmids ) were then polished by mapping raw sequencing reads back onto the assembled circular contigs . 
The detection of methylated bases and clustering of modified sites to identify methylation-associated motifs was performed as previously reported ( 68 ) . 
In brief , raw reads were aligned to the complete genome of MS6198 , and interpulse duration ( IPD ) ratios were calculated using PacBio 's in silico kinetic reference computational model . 
Sequence analysis , annotation , and in silico typing . 
Visualization and annotation of plasmid sequences were performed using PROKKA v1 .11 ( 69 ) and the Artemis Genome Browser ( 70 ) . 
Sequence comparisons were constructed using WebACT ( 71 ) and visualized with Easyfig 2.1 ( 72 ) and Artemis Comparison Tool ( ACT ) ( 73 ) . 
In silico DNA manipulations and analysis were conducted and visualized in CLC Main Workbench ( version 7.0.2 ; Qiagen Bioinformatics ) and Easyfig 2.1 ( 72 ) . 
In silico E. coli multilocus sequence typing was performed using the MLST database hosted at the University of Warwick ( 74 ) . 
Plasmid Inc types were determined by PCR-based replicon typing ( 63 , 64 ) or in silico using PlasmidFinder ( 60 ) . 
Collection of IncA/C complete sequences and analysis . 
Complete sequences of IncA/C plasmids from GenBank were selected by BLASTn using the sequence of RA1 repA as a reference ( 90 % identity and 90 % query coverage ) . 
The BLASTn hits were manually reviewed , and only published sequences were included in our IncA/C plasmid database . 
Each plasmid sequence was also verified by BLASTn to contain IncA/C replicon-typing primer binding sites ( 63 ) . 
With the addition of pMS6198A , this database comprised 82 IncA/C plasmid sequences ( as of 9 May 2016 ) ( see Data Set S5 in the supplemental material ) . 
The gene annotations of pMS6198A were used as a reference to identify genes present in all 82 IncA/C plasmids , using default BLASTn v2 .2.26 settings with the criteria of an expected value of 10 30 and minimum coverage of 95 % . 
The sequence of each conserved gene was extracted from each plasmid using EMBOSS v6 .5.7 ( 75 ) . 
PMLST minimal spanning trees were built by Phyloviz using the goeBURST algorithm ( 76 ) . 
All alignments were constructed in MEGA 6.06 ( 77 ) using ClustalW with default settings . 
Phylogenetic trees were produced in MEGA 6.06 using maximum likelihood with default settings and supported with 1,000 bootstraps . 
The presence or absence of resistance genes was determined using the BLASTn algorithm ( 100 % identity at 100 % coverage ) against the resistance gene database ARG-ANNOT ( 78 ) . 
IncA/C sequence analysis and metadata have been incorporated into the microreact database ( https://microreact.org/project/IncACPlasmids?tt rc & tns 4 & tts 6 ) ( 80 ) . 
Accession number ( s ) . 
The sequences for the MS6198 chromosome , pMS6198A , pMS6198B , pMS6198C , and pMS6198D have been deposited in the NCBI GenBank database under accession numbers CP015834 to CP015838 , respectively . 
Raw PacBio sequence reads for MS6198 and Illumina MiSeq reads for duplicate TraDIS runs have been deposited in the Sequence Read Archive ( SRA ) under accession numbers SRX1797306 , SRX1992326 ( replicate 1 ) , and SRX1992327 ( replicate 2 ) , respectively . 
SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/ AAC.01740-16 . 
ACKNOWLEDGMENTS
We thank David Miller , Tim Bruxner , and Angelika Christ from the Queensland Centre for Medical Genomics ( Institute for Molecular Bioscience , University of Queensland ) for technical support with the TraDIS protocol . 
The protocol was set up in consultation with Brian Fritz ( Illumina ) and Sabine Eckert , Daniel Turner , and Matthew Mayho ( Wellcome Trust Sanger Institute ) . 
This work was supported by grants from the National Health and Medical Research Council ( NHMRC ) of Australia ( GNT1033799 and GNT1067455 ) and High Impact Research ( HIR ) grants from the University of Malaya ( UM-MOHE HIR grant UM C/625/1 / HIR/MOHE/CHAN / 14/1 , no . 
H-50001-A000027 ; UM-MOHE HIR grant UM C/625/1 / HIR / MOHE/CHAN/01 , no . 
A000001-50001 ) . 
M.A.S. is supported by an NHMRC Senior Research Fellowship ( GNT1106930 ) , and S.A.B. is supported by an NHMRC Career Development Fellowship ( GNT1090456 ) . 
Teik Min Chong is supported by the Postgraduate Research Fund ( PPP ) ( grant no . 
PG080-2015B ) .