The Pandoraea species isolates P. faecigallinarum DSM 23572^T and P. apista DSM 16535^T were obtained from the Leibniz Institute DSMZ—German Collection of Microorganisms and Cell Cultures GmbH. The isolates were cultured on Luria-Bertani medium for 24 h at 28 °C with shaking at 220 rpm in reference to a previous protocol (Lim et al., 2014). Genomic DNA was isolated using an Epicentre Masterpure DNA extraction kit (Epicentre, Inc., Madison, WI, USA) according to the manufacturer's instructions. To obtain complete genome sequences for a better genomic overview, Pacific Biosciences (PacBio) RS II Single Molecule Real Time (SMRT) sequencing method (Pacific Biosciences, Menlo Park, CA) was employed. Genomic DNA was processed into SMRTbell™ library using the SMRTbell template preparation kit 1.0 according to the “Procedure and checklist-20 kb template preparation using BluePippin size-selection system” protocol as reported by Lim et al. (2016). Sequencing of the SMRTbell™ libraries were conducted using P6 DNA polymerase (C4 chemistry) which is PacBio's latest DNA sequencing chemistry that improves the average read length (http://investor.pacificbiosciences.com/releasedetail.cfm?releaseid=876252). Quality filtering and de novo assembly of the sequenced reads were performed using the Hierarchical Genome Assembly Process (HGAP) version 2 of PacBio (DevNet; Pacific Biosciences).

Based on the PacBio Circularizing and Trimming wiki at (https://github.com/PacificBiosciences/Bioinformatics-Training/wiki/Circularizing-and-trimming), circularity of all contigs generated from the assembly was determined based on reads mapping across the beginning and end of the contigs. To confirm this, Gepard (Krumsiek et al., 2007) was used to generate dot plots for each contig to detect the presence of self-similar region at the 5′ and 3′ contig ends. The contigs were then subjected to PacBio Quiver algorithm for polishing prior to further analysis. Chromosomal contigs were determined based on the similarity of the contig sizes to the expected chromosomal size of Pandoraea and were further confirmed by using BLASTN. Subsequently, all extra-chromosomal contigs were aligned against the chromosomal contigs and were assessed using Contiguity (Sullivan et al., 2015). Only contigs which demonstrated low or no similarity to their respective chromosomal contigs were determined as plasmid contigs.

As a result, three plasmid contigs were observed. Non-clinical isolate P. faecigallinarum DSM 23572^T harbors two plasmids (pPF72-1, pPF72-2) whereas clinical isolate P. apista DSM 16535^T carries a single plasmid (pPA35). Complete sequences of plasmids from non-clinical strains P. oxalativorans DSM 23570^T (pPO70-1, pPO70-2, pPO70-3, pPO70-4) (Chan et al., 2016) and P. vervacti NS15 (pPV15) (Ee et al., 2015) were obtained from NCBI.

Prokaryotic Dynamic Programming Genefinding Algorithm (Prodigal) Version 2.6.2 (Hyatt et al., 2010) was used in the prediction of open reading frames. The predicted sequences were then functionally annotated by performing homology search against the NCBI nr database, InterProScan (Quevillon et al., 2005) and the Conserved Domain Database (CDD) (Marchler-Bauer et al., 2015). Identification of key genes conferring biological advantages were performed based on Virulence Factors Database (VFDB) (Chen et al., 2005), The Comprehensive Antibiotic Resistance Database (CARD) (McArthur et al., 2013) and TAfinder (http://202.120.12.133/TAfinder/index.php).

The accession numbers of the complete sequences of the three plasmids sequenced in this study and of the five plasmids from NCBI can be found in GenBank with the accession numbers CP011808 (pPF72-1), CP011809 (pPF72-2), CP013482 (pPA35), CP011518 (pPO70-1), CP011519 (pPO70-2), CP011520 (pPO70-3), CP011521 (pPO70-4), and CP010898 (pPV15).

All eight plasmids range in sizes from 46,278 to 640,227 bp, organized in a single contig with overall G + C content in the range of 57.81–63.8%. The number of open reading frames in the range of 65–626 was predicted across the plasmids in which 62–591 are protein-encoding genes (Supplementary Figures S1–S4 and Supplementary Table S1). The features of each plasmid are shown in Table 1. The putative origin of vegetative replication (oriV) and terminus (terV) for each plasmid were identified based on cumulative GC-skew analysis using GenSkew (http://genskew.csb.univie.ac.at/) (Supplementary Figures S5–S8). The predicted oriV and terV are marked by the minimum and maximum values of the cumulative GC-skew, respectively, which corresponds to the nucleotide positions (Grigoriev, 1998). Overall, Pandoraea plasmid sequences yielded low similarity against plasmid sequences from the RefSeq and EMBL nucleotide sequence database with sequence completeness of <12%. In addition, no homologous gene organization was observed. Hence, no comparison between regions of the Pandoraea plasmids and other known plasmids could be made. These findings indicate that the Pandoraea plasmid sequences were not reported before.

To determine the possible origins of genes within the Pandoraea plasmids, we performed a search against a non-plasmids database. The database was constructed based on bacterial genome sequences obtained from NCBI. We observed that majority of the genes across the plasmids matched to genomes of Burkholderia, Bordetella, Xanthomonas, Ralstonia, Pseudomonas, Alicycliphilus, Nitrosomonas and Cupriavidus with similarity and completeness scores well above 50%.

Plasmid replication is vital in ensuring reproducibility and segregation of plasmids into daughter cells during host cell division. This is an essential aspect in the plasmid maintenance system and highlights the values of identifying the potential genes involved in replication machinery, especially in the unknown potential of plasmids in Pandoraea.

Putative replication initiator protein-encoding gene (repA) was identified in pPO70-4 and pPA35, indicating self-replication potential of these two plasmids. However, no repA was identified in the other plasmids. This has led us to speculate that a different replication mechanism exists in which a plasmid initiator protein is not employed, such as the case for plasmid ColE1 (del Solar et al., 1998) that reportedly uses DNA polymerase I for replication initiation.

Putative genes surrounding the putative predicted oriV and terV locations that could be involved in replication were identified. In pPA35, oriV was identified at position 72,001 bp within an acyltransferase gene whereas terV was identified at position 31,001 bp within an intergenic region between conjugal transfer TraM containing domain protein and hypothetical protein-encoding genes. Conjugation tra and trb genes were found to flank either side of the terV. A similar observation was made in plasmid pRF in which its terV was flanked by conjugation genes (Gillespie et al., 2007). Other genes located downstream of the terV in pPA35 are partitioning genes (parA, parB) and repA. A single-strand DNA-binding (SSB) protein-encoding gene (ssb) was found located next to repA and could be associated with DNA replication (Meyer and Laine, 1990).

The oriV at position 37,001 bp in pPO70-4 was found within an antitoxin gene. Located upstream are genes repA, ssb and a gene encoding for a partitioning protein which shares 96% identity with ParA of Xanthomonas. Another partitioning protein with a ParA domain was found downstream of the oriV and has 99% identity to the upstream partitioning protein-encoding gene. At position 25,001 bp, the terV is located within a type IV secretion system (T4SS) protein-encoding gene flanked on either side by conjugation genes. This is another feature similar to plasmid pRF (Gillespie et al., 2007).

As for pPF72-1, the oriV at position 250,000 bp is found within an intergenic region between parA and a hypothetical protein-encoding gene, a situation that matches the finding that parA is often located adjacent to oriV (Picardeau et al., 2000). The terV at position 332,001 bp is located within a hypothetical protein-encoding gene. Plasmid pPF72-2 on the other hand, has its oriV at position 124,001 bp in an intergenic region between transposase and hypothetical protein-encoding genes whereas its terV at 30,001 bp is found within a hypothetical protein-encoding gene with ssb located downstream.

Overall, only pPA35, pPF72-1, pPF72-2 and pPO70-4 showed the presence of genes that are possibly involved in their replication systems. Plasmids pPO70-1, pPO70-2, pPO70-3 and pPV15 showed no signs of replication genes located near the oriV and terV as well as in other parts of the plasmid. The observed characteristics raised the hypothesis of plasmid-host interaction for plasmid replication, as reported for plasmid Pps10 from Pseudomonas syringae (Fernández-Tresguerres et al., 1995). The absence of replication genes in the four plasmids motivates the understanding of the interaction between the plasmids and their respective host strains. The information can be potentially used to design the plasmids with a host-killing system, and enable them to be released into the environment as an effort to control possible infection by Pandoraea and other opportunistic pathogens.

Partitioning genes function in the segregation of plasmids into daughter cells following plasmid replication. The segregation activity, that takes place as part of the second mechanism of plasmid maintenance, is also known as the partition system or par system. In the identification of putative partitioning genes, genes parA and parB of the type I par system (Gerdes et al., 2000) were observed in pPF72-1 and pPA35. These genes are trans-acting protein-encoding genes whereby parA encodes an ATPase that regulates the function of DNA binding protein-encoding parB (Bignell and Thomas, 2001). Besides the two aforementioned plasmids, pPO70-4 was found to contain two partitioning protein-encoding genes with a ParA domain identified in each gene.

No par genes were identified in pPF72-2, pPO70-1, pPO70-2, pPO70-3 and pPV15, indicating a possibly different mode of segregation besides the typical partition system. This situation is possible considering previous studies conducted on plasmids that describe the usage of other approaches for segregation other than the par system. One of these studies involves plasmid pSK1 from Staphylococcus aureus, in which an identified gene is involved in the increase of segregation stability while no ATPase activity was required to drive plasmid segregation (Simpson et al., 2003). In another study, plasmid R388 utilizes other genes instead of the par system for plasmid segregation (Guynet et al., 2011).

Interestingly, all the Pandoraea plasmids without par genes also did not contain repA. The consistent observations on this absence of replication and partitioning genes, strengthen the view on close interactions between the host strains and plasmids for the maintenance system. Plasmid stabilization/stability protein-encoding genes have been observed across pPF72-1, pPF72-2, pPO70-2, pPO70-3 and pPV15, possibly playing a part in plasmid segregation but this has not been confirmed.

Toxin-antitoxin (TA) systems in plasmids form the third mechanism of plasmid maintenance. They have also been termed as post-segregational killing systems or plasmid addiction systems (Kroll et al., 2010). A single TA system consists of an antitoxin and a toxin protein. In the event that a daughter cell inherits a plasmid, the antitoxin protein acts to neutralize the action of the toxin, thereby preventing cell death (Sengupta and Austin, 2011). However, when plasmids are not distributed to daughter cells, the stable toxin protein causes cell death whereas the unstable antitoxin protein is degraded. In other words, TA systems prolong the presence of plasmids in the bacterial population by eliminating plasmid-free cells that have arisen due to unsuccessful plasmid replication and segregation. There are five classes of TA systems of which the class Type II is the best studied (Unterholzner et al., 2013). Hence, we have focused only on class Type II TA systems in our study.

The presence of putative genes known to behave as TA systems, have been identified across the Pandoraea plasmids except for pPV15. Supplementary Table S2 displays the list of TA systems found in each of the plasmids.

More than one TA system was observed in each of the plasmids except in pPO70-4. The observed presence of TA systems could be due to the nature of TA as a mobile element, and is subsequently integrated into the DNA of the plasmids. The presence of TA systems would enhance the post-segregational activity, and in turn ensure better plasmid maintenance. With the exception of pPO70-4, all plasmids in free-living strains have more TA systems as compared to pPA35, which originates from a host-associated strain. A study carried out by Pandey and Gerdes (2005) presented that free-living organisms possess multiple TA systems as compared to host-associated organisms that did not have any. This observation is similar to the Pandoraea plasmids from free-living strains, whereby the presence of TA systems proposes the need for survival in the constantly changing ecosystem. However, the identification of two TA systems in pPA35, suggests that the maintenance of this host-associated strain-harboring plasmid is still essential in adapting to an inconsistent environment.

A total of seven types of TA systems were observed across the Pandoraea plasmids namely MazEF, VapBC, RelBE, YgiT-MqsR, HigBA and ParDE. The TA systems VapBC and RelBE of pPF72-1 and pPO70-3, respectively, were found within plasmid stabilization protein-encoding genes. It is proposed that these TA systems are helpful in preventing the loss of these elements in daughter cells. All plasmids except pPO70-2 and pPO70-4, have more than one copy of some of its TA systems. Such a situation will ensure a stable single-copy plasmid inheritance (Venkatesan et al., 2001). All TA systems identified in the Pandoraea plasmids were encoded near transposase or integrase encoding genes, indicating an ability to be mobilized by mobile genetic elements (Bustamante et al., 2014).

The orientation of type II TA systems is such that the antitoxin gene is situated upstream of the toxin gene (Van Melderen and Saavedra De Bast, 2009). However, we observed that some TA systems in the Pandoraea plasmids have an opposite orientation. This was observed amongst MazEF, VapBC, RelBE, HigBA and HipBA across pPF72-1, pPF72-2, pPO70-1 and pPO70-3, whereby RelBE appears to be the most frequent in number having an opposite orientation. Previous studies have also shown the same unusual occurrence in TA system orientation (Budde et al., 2007; Yamaguchi et al., 2009; Christensen-Dalsgaard et al., 2010).

Gene overlap in the TA systems has been observed across the Pandoraea plasmids. The occurrence of gene overlap in prokaryotes indicates translational coupling and this phenomenon is common in TA systems (Pandey and Gerdes, 2005). An overlap of 4 and 1 bp in TA systems were observed in pPF72-1, pPF72-2, pPO70-1, pPO70-2, pPO70-4 and pPA35. Generally, most TA systems have an overlap of 4 bp or even 1 bp, in which the antitoxin stop codon overlaps the start codon of the toxin gene (Pandey and Gerdes, 2005). Other overlap sizes have also been observed amongst the Pandoraea plasmids. Overlaps of 7 and 5 bp were found within MazEF and RelBE, respectively, in pPF72-1. Plasmid pPF72-2 has 29, 11, and 20 bp overlaps in its RelBE whereas a 9 bp overlap is seen in HipBA in pPO70-1. Lastly, 13 and 29 bp overlaps were identified in RelBE in pPO70-3. Chan et al. (2012) have previously reported on other overlap sizes besides the usual 4 and 1 bp overlaps. Some TA systems in the Pandoraea plasmids have no gene overlap such as in pPF72-1, in which gaps of 118 bp (MazFike_domain, higA) and 47 bp (RelBE) were observed whereas a gap of 170 bp was identified within RelBE in pPO70-3.

Conjugation is one of the mechanisms of bacterial lateral gene transfer (Burrus and Waldor, 2004). Genes enabling conjugation are often located on plasmids and genetic transfer occurs when these conjugative plasmids are self-transmitted during bacterial cell to cell contact. The mode of conjugation is important as it enables plasmids to deliver features that can be beneficial to the host strains. An example has been reported by Rösch et al. (2014) where the presence of conjugative plasmid pLS20 conferred stress resistance to its host strain. The identification of the conjugation mechanism can also be helpful in preventing the transfer of pathogenic properties such as reported by Lujan et al. (2007), who described the disruption of relaxase to prevent antibiotic resistance transfer. The conjugative potential of Pandoraea plasmids has been predicted based on the presence of putative conjugation genes encoding T4SS components. Overall, these genes were predicted as tra, trb, trw and virB. These different gene designations exist overtime for different plasmids in particularly when they are categorized into different incompatibility groups (Stolz, 2014). Some of the examples of this categorization include tra, trb and trw of incompatibility groups IncF and IncN, IncP and IncW, respectively, as well as virB of Ti plasmids (Christie, 1997). All these gene designations have been identified in the conjugation genes among the Pandoraea plasmids. No conjugation genes however, were identified in pPO70-3.

Conjugation genes in pPA35 and pPO70-4 were predicted as tra, trb and virB. In pPA35, two separate regions exist mainly coding for tra (tra region) and trb (trb region) genes, respectively. Similar regions were observed in plasmid pTiC58 of Agrobacterium tumefaciens (Alt-Mörbe et al., 1996). The tra region contains genes traC, traG, traJ and traL, in which all except traL are located on the lagging strand. Genes encoding a relaxase and peptidase S26, were also identified in this region in which the latter contains conjugal transfer peptidase TraF domain. An additional gene adjacent to the tra region was characterized with conjugal transfer protein TraM domain and is located on the leading strand. Genes in the trb region are trbB, trbC, trbD, trbG, trbI, trbJ, trbL and traX. A single virB4 has also been identified in this region. Five additional genes were observed within and adjacent to the trb region with domains TrbA, TrbF, TrbH, TrbM and TrbN, respectively. For the trb region, the conjugation and additional genes were located on the leading strand except for the additional gene with conjugal transfer protein TrbA domain.

In contrast to the two conjugative regions in pPA35, pPO70-4 comprises a single conjugative region which contains genes traC, trbL, traB, virB10 and virB3. Besides these, genes encoding pilus assembly protein CpaF and T4SS protein were also identified. Three additional genes were identified within and adjacent to the conjugative region of pPO70-4. These genes have domains TrbG, TrbF and T4SS lytic transglycosylase VirB1, respectively. Conjugation genes in pPO70-4 are all located on the lagging strand except for traC.

The best hit was identified for each additional gene via homology search with BLASTP against the nr database. Information on their domain and best hits are tabulated in Supplementary Table S3. They were found to carry conjugation domains and have close hits to conjugation genes suggesting that they are yet to be reported as conjugative-related genes.

In comparison to the conjugation genes in pPA35 and pPO70-4, the other Pandoraea plasmids contain differently annotated conjugation genes, except for a single Tra coupling protein-encoding gene (traD) found in pPO70-1, pPO70-2 and pPV15. These genes include trwB, virB11 and virB9. Relaxase, endonuclease and lytic transglycosylase and type VI secretion system (T6SS) protein-encoding genes were also identified.

Domain search revealed a T6SS lytic transglycosylase VirB1 domain within the lytic transglycosylase encoding genes, indicating that they could function in degrading the peptidoglycan wall in order to permit construction of structures such as the pilus during conjugation (Wallden et al., 2010). Furthermore, these genes have 59-65% identity with T4SS protein VirB1 of Xanthomonas fuscans subsp. Aurantifolii str. ICPB 11122. According to Wang and Macrina (1995), the endonuclease, also known as the nickase, functions in carrying out a nick at the oriT during conjugation. It is suggestive that the endonuclease gene identified in the Pandoraea plasmids, may act in a similar way. The endonuclease genes found in pPF72-2, pPO70-1 and pPV15, share similarity to the plasmid conjugative transfer endonuclease encoding gene of Collimonas arenae (60% identity), Burkholderia sp. AU4i (69% identity) and Burkholderia cenocepacia H111 (69% identity), respectively.

The conjugation region in pPF72-1, pPF72-2, pPO70-1, pPO70-2 and pPV15 is presented in Figure 1 as the only significant region that is shared amongst the Pandoraea plasmids. The shared region contains the same conjugation genes except for traD, which is only present in pPO70-1, pPO70-2 and pPV15. Gene arrangements in this region appear to be conserved amongst the five plasmids. Pairwise comparisons between the conjugation regions yielded 88–100% nucleotide identity values shared between the respective genes, further leading to the possibility of common origin for these regions.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.