All recorded QacC proteins encoded by RC-plasmids are strictly conserved. Seventy GenBank entries describe identical proteins, sequenced from different staphylococcal plasmids. Only two amino acid (aa) substitutions have thus far been described: an A → S substitution at aa position 9 has been recorded five times (four from Staphylococcus warneri plasmids and one from a non-speciated Staphylococcus spp.) and a change of I → T at position 30 has been found twice in S. epidermidis plasmids.

Surprisingly, the conservation of qacC at the nucleotide (nt) level is also nearly absolute: in addition to the two mutations resulting in the aa substitutions, only one synonymous mutation has been encountered so far (an A → G mutation at nt position 81 in pNVH99 of S. aureus). Due to this strict conservation, which is indicative of a recent bottle-neck selection event, phylogenetic analysis of the QacC protein or its gene is not possible, whereas inclusion of the sequences flanking qacC introduces informative variation. Analysis of a number of complete qacC-bearing RC-plasmid sequences (selected for their length of approximately 3 kb) reveals that these are variable, apart from their strictly conserved qacC gene plus its flanks, as shown in Figure 1.

The plasmids used in Figure 1 fall into two types depending on their rep gene, which we define as Type I and Type II, respectively, as summarized in Table 1. Rep1, the determinant for Type I plasmids, is typically 282–307 aa long while Rep2, indicative of Type II, is 332–334 aa. Rep2 homologs were previously described as RepNVH99 (Bjorland et al., 2001) and REP416 (Alam et al., 2003). Both types of plasmids mostly lack other recognized genes and their rep is found invariably upstream of qacC on the opposite strand, so that the two genes are divergently transcribed.

Even within their types, Rep1 and Rep2 are still less conserved than QacC: Rep1 is conserved for 77% at the aa level (68% at nt level), and for Rep2 these percentages are 82 and 78%, respectively. The difference in nt conservation between rep and qacC dictates these genes did not co-evolve on the RC-plasmids where they now reside. Instead, the observation can only be explained by assuming that qacC was recently incorporated into plasmids with variable rep genes, an event that occurred so recently that third-base variation has not yet been introduced.

In an attempt to explain the finding of a conserved qacC in these variable backgrounds we compared the flanking sequences of qacC in the RC-plasmids listed in Table 1. Key features of the multiple DNA alignment are summarized in Figure 2.

Upstream of qacC 162 nucleotides are identical in Type I and Type II plasmids except for two regions: a variable number of direct repeats (DR1) is present toward the distal end of the upstream flank, and an 8 nt repeat insert is found in two Type II plasmids, in a region where a qacC' insert is found in a number of other plasmids, which will be revisited in the next section. The distal end of this 162-long upstream region starts with the DSO, the nature of which will be discussed at the end of this paper. Beyond the DSO, moving further away from qacC in the upstream direction (toward the start of the replication gene), the sequences diverge between Types I and II (Figure 2). Downstream of qacC, Type I and Type II plasmids share 164 nt before their sequences diverge. The exception is pSH416, which contains a large insert, although it cannot be excluded that this is the result of an assembly error.

Thus, the borders of a conserved qacC locus shared between Types I and II plasmids is defined from 162 nt upstream (bearing a DSO element) to 164 nt downstream of the qacC open reading frame, and includes an SSO element. We proposed this “qacC element” has recently been spread over short RC-plasmids.

As has been previously described, a number of short RC-plasmids contain a repeat of 142 nt, composed of the first 97 coding nt of qacC, together with 45 upstream nt (indicated by a gray box in Figure 2). The fragmented open reading frame is frequently annotated as qacC' (previously called Delta-qacC) but the gene fragment is too short to produce a functionally active protein (Leelaporn et al., 1995). Phenotypic evidence from pSM52 (Costa et al., 2013) or pSW49 (Bjorland et al., 2006) suggests that the promoter of qacC is present on the 45 nucleotides that separate the qacC' fragment from the active, complete gene (the promoter is also present in front of qacC').

Plasmids containing the 142-nt duplication of qacC' are designated Type Ia or Type IIa, depending on their Rep (Table 1). That this partial duplication can be variably present on both types can be explained in two ways. Either the qacC locus was distributed to Type I and Type II plasmids first, after which an identical qacC' duplication occurred independently in both, or the duplication occurred first and the locus with or without the duplication spread independently in both types of plasmids. Analysis of the downstream sequences provides a clue as to which scenario is more likely. A single-nucleotide insertion is found 137 nt downstream of qacC in all known Type Ia and Type IIa loci (indicated by an asterisk in Figure 2), and it is also present in some Type II loci (e.g., pSA1308 or pLUH01, see Table 1). It is therefore proposed that the duplication occurred in such a Type II plasmid (Figure 3), after which the whole locus including the duplication was distributed over Type I plasmids. This observation supports the hypothesis that the QacC locus is transferable.

A few Type Ia plasmids contain multiple rep (pseudo)genes, for example pSW49 of S. warneri (Bjorland et al., 2006). This plasmid of 3552 bp length contains, besides qacC', qacC, and rep1, a truncated rep2 pseudogene. Although occasionally RC-plasmids are encountered that contain two different but active replication proteins (e.g., pMRI5.2 of Lactobacillus plantarum, Cho et al., 2013), in most cases only one functional replication initiation protein is present. Plasmid pSW49 is possibly a fusion product of a Type Ia and a Type IIa plasmid, in which only one copy of the qacC'-qacC locus was maintained, while Rep2 was truncated. Likewise, S. warneri plasmid pSW174 is likely a fusion plasmid that in this case retained both Rep genes.

Alam et al. (2003) defined a “Type 3” class based on the plasmid of S. aureus strain SH651. This unnamed and incompletely sequenced plasmid (called pSH651 in Table 1) contains an insertion sequence directly downstream of qacC. It further contains a 122 nt direct repeat upstream of qacC. The total length of this plasmid is not recorded and further sequences have not been made available, so the exact nature of this plasmid cannot be determined at this stage. However, the 122 nt repeat is part of a sequence that is also repeated in other RC-plasmids bearing qacC, which we call Type III.

The prototype of Type III plasmids is pST827 (Table 1), sequenced from a strain of unspecified “Staphylococcus sp.” isolated from food (Heir et al., 1995) (an incorrectly assembled version of this plasmid appeared in Alam et al., 2003). In Type III plasmids, qacC is flanked by two identical 229 nt-long sequences so that the upstream and downstream flanks of qacC are identical, except for one mismatch (an extra T present in the downstream unit). For this reason, Type III plasmids do not match the conservation of the upstream qacC flank (Figure 2) and were excluded from the homology plot in Figure 1.

Two models can be proposed how the qac locus of Type III plasmids relates to Types I and II, shown in Figure 4. The simplest model (Figure 4A) proposes that Type III is formed from a Type I or II qacC plasmid by duplication of the downstream flank inserted into the upstream flank, at position −45 nt. Alternatively (Figure 4B), Types I and II are formed from Type III, which requires a hypothesized precursor in which upstream sequences have been duplicated into the downstream flank. Although that model requires two steps, it may be the more likely scenario, because in the hypothesized precursor a DSO and SSO would be separated by 316 nt, a distance that is more typical for staphylococcal RC-plasmids that only contain a rep gene (e.g., pSK6 of S. aureus, acc. nr. NC_001995).

Table 1 lists pSP187 from Staphylococcus pasteuri as a second example of a Type III RC-plasmid. It contains a different replication gene as well as a few other genes—it may have resulted from a recombination with another plasmid. So far, pST827 and pSP187 are the only completely sequenced short RC-plasmids containing a Type III qacC.

The qac locus of Type III RC-plasmids can also be found on long conjugative plasmids; seven of these (Table 1) were used for comparative analysis (a graphical representation is given in Figure S1). In these plasmids two transposase genes that are part of Insertion Sequence IS257 copies border the locus. In between these ISs a rep1 gene is present. This suggests that a complete Type I plasmid was incorporated into a conjugative plasmid, after which recombination events duplicated the SSO-bearing fragment and deleted the DSO flank immobilized the plasmid. The public database does not contain short qacC RC-plasmids of Type III that contain rep1.

We noted that the insertion of the duplication observed in Type III plasmids occurred at the same position as the insertion of qacC' in Types Ia/IIa. This suggests the −45 position is frequently subject to recombination. Indeed, in two Type II plasmids (pWBG754 and pNVH99), 18 nt are repeated here, replacing 10 nt of original sequence, thus resulting in an insertion of 8 nt in the alignment (Figure 2). That the position 45 nt upstream of qacC is a frequent target for recombination may be due to the local physical properties of the sequence. An analysis of the DNA structural properties revealed this is an AT-rich region that is curved and will easily melt.

Relatively few sequences are available for Qac members of the SMR family other than QacC. Less than 10 records each are currently available for QacG, QacJ, or QacH, and a number of these are contigs from genome or metagenomic sequencing projects. All these plasmids contain a rep2 homolog.

The nucleotide sequences of five QacG plasmids (four RC-plasmids and one segment of a conjugative plasmid) listed in Table 2 produce a similar conservation pattern as observed for QacC plasmids: the open reading frame of qacG is 99% conserved (5 mismatches only) whereas rep2 is only 76% conserved. Again, the nearly complete conservation of qacG is extended to its flanking sequences, encompassing the DSO and SSO, although the homology is continued 22 nt upstream of the DSO nick site. In contrast, the five analyzed QacJ sequences (three from contigs) are stronger conserved for rep2 (91%) than for qacJ (83%) and the homology of the rest of the plasmids suggests they are closely related (data not shown). These observations show that there is no functional selection that would prevent SMR-type qac genes from accumulating nt mutations: qacG seems to have been around long enough to collect a limited number of mutations already, in contrast to qacC. There are not enough completely sequenced qacH plasmids publically available to assess their homology.

The homology between plasmids bearing qacC (Type II), qacG, qacJ, and qacH was used to identify conserved regions (Figure 5). These were restricted to the DSO, a short region immediately downstream of qac, and the distal end of the SSO region.

Adding sequences of Type I QacC plasmids to the alignment on which Figure 5 was based identified conserved sequences essential for replication, summarized in Figure 6. The distal end of the 165 nt-long QacC upstream flank including the DSO is strictly conserved, though the third base is G instead of T in Type I QacC plasmids only. The direct repeat unit DR1 of QacC plasmids is also present in all QacG plasmids, two QacJ plasmids and in the single QacH plasmid that was analyzed. Besides in QacC plasmids, the DR2 unit is repeated in two QacJ plasmids (the three other QacJ sequences contain an alternative DR2). In all plasmids, the DSO is followed by a stretch of low complexity sequences containing multiple short homo-nucleotide repeats. The rest of the upstream flank is not conserved between these Qac plasmids.

The downstream flank of all qac genes is more strongly conserved than the upstream flank. It starts with a 12 bp-long direct repeat, (AATTGCTTTATT, indicated as DR3 in Figure 6), which is completely conserved in case of qacC and qacJ; the first unit is mutated in qacG and both units contain mutations in qacH sequences. The very distal end of the downstream flank contains 18 nt that are strongly conserved (qacJ contains one mismatch in the last T). Beyond this, homology is lost. Weakly conserved inverted repeats are present in the downstream flank, which can form a hairpin structure (Figure 6B) related to the function of SSO.

The conserved sequence of the DSO given in the box of Figure 6 is the strand on which the qac gene is found. The arrow indicates the position where Rep introduces a nick in the complementary strand, which initiates replication of the plus strand (Gruss et al., 1987; Khan, 2005; Bikard et al., 2010). RC-plasmids are divided into different families depending on the type of replication protein and the DSO sequence this protein recognizes. Originally, four eubacterial RC-plasmid families were recognized (del Solar et al., 1998) which were extended to over 12 (Khan, 2005) and now 17 (Ruiz-Masó et al., 2015) plasmid families. Based on aa similarity, Rep1 of Type I plasmids belongs to the pC194-family (Seery et al., 1993), a representative of the incB incompatibility group (Thomas et al., 1990). The conserved DSO resembles that of pC194 with only one mismatch and one insertion (the DSO of pC194 is TATTAT↓CAAGATAAGAAAGA, Bikard et al., 2010). The nature of Rep2 is less clear, though it is more closely related to Rep of pC194 than to Rep proteins of the other families; information on its inc group could not be identified. Rep2 recognizes the same DSO as Rep1 with the exception of the polymeric nucleotide positioned four bases upstream of the nick site, which is G instead of T.

In a number of well-studied RC-plasmids, the nick site is positioned in the loop of a stem-loop structure that is formed by inverted (palindromic) repeats (Ruiz-Masó et al., 2015). However, from the DSO and sequences surrounding the nick site in Qac plasmids inverted repeats are absent. Lack of hairpin structures in a DSO has been observed before in other members of the pC194-family, for instance in pUB110 of Bacillus subtilis (Alonso et al., 1988). A functional DSO further depends on a binding element for the Rep dimer. This binding element is usually less conserved than the nick site and in case of the pC194 family typically contains inverted repeats (del Solar et al., 1998; Bikard et al., 2010). Since these are absent in case of Qac, we propose that two sets of direct repeats following the nick site function as the binding region for Rep1 and Rep2.

The conserved SSO found in the downstream flank of qacC belongs to the ssoA-type (Kramer et al., 1998). Irrespective of whether Qac is present or absent, this particular SSO sequence is almost exclusively found in Staphylococcus plasmids, in accordance to experimental evidence that SSO sequences are host-specific (Kramer et al., 1998), although promiscuous within the genus.

In the Type III QacC plasmid pST187, where the SSO flank is duplicated, the event has replaced the DSO. Lack of a recognizable DSO element has been pointed out before (Bjorland et al., 2007). These authors proposed an alternative DSO element, and in combination with its unique Rep gene it was suggested that pSP187 belongs to RRC-group VI plasmids (Bjorland et al., 2007).

The fact that Qac is positioned in between SSO and DSO on RC-plasmids may explain how it is mobilized. During rolling circle replication, the plus strand is formed from the nick site, with the replication fork moving all the way round toward the SSO, to continue beyond the original nick site. As Khan describes: “Once the replication fork reaches the termination site, i.e., the regenerated DSO, DNA synthesis proceeds to approximately 10 nucleotides beyond the nick site. (…) The SSO [is] generally located immediately upstream of the DSO such that [it is] not exposed in a single-strand form until the leading strand has been almost fully synthesized” (Khan, 2005). Initiation of replication is tightly regulated, and how re-initiation is inhibited until the cycle is completed is well understood for some RC-plasmid families (e.g., pT181, reviewed in Ruiz-Masó et al., 2015), but for the pC194 family, to which Qac RC-plasmids belong, the regulation is not clear (Ruiz-Masó et al., 2015). The presence of a gene (qacC) in between SSO and DSO is an anomaly for these plasmids, as the sso is usually located close to the dso (Ruiz-Masó et al., 2015). During replication, which starts from DSO (with qacC being replicated last) an intermediate stage exists when the single-strand positive strand is covalently bound to Rep at the nick site, to be recircularized when Rep is released. This circular ssDNA is completed to dsDNA starting at the SSO hairpin, again with qacC being replicated last, as shown in Figure 7. If, however, a new nick is introduced in DSO before the minus strand replication round is completely finished, a nicked DNA fragment spanning from the nick site to the SSO, with qacC in between, would be formed. This could be incorporated in an acceptor plasmid by homologous recombination, making use of host recombination proteins. Untimely initiation of replication would normally be prevented when SSO and DSO are close together, but it is quite possible that the presence of a qac gene in between these elements upsets the timing of events.

In search of support for our model, we looked for other genes located between SSO and DSO of RC-plasmids. In species other than Staphylococcus this is quite common for rep or mob genes, and a variety of other genes can be found to separate the two replication elements (e.g., Alonso et al., 1988; Andrup et al., 2003; Alegre et al., 2005; Wu et al., 2007; Zhang et al., 2010, 2011). None of these genes in RC-plasmids from genera other than Staphylococcus share significantly higher homology compared to the rest of their plasmid background, thus there is no indication of gene transfer. Whether other genes on RC-plasmids from genera other than Staphylococcus use this mode of mobilization is currently not known.

However, a second example exists of a gene likely to be transferred by the same mechanism. The gene lnuA, resulting in lincosamide resistance, is located in between SSO and DSO on RC-plasmids of Staphylococcus species, and again it is stronger conserved at the nt level than the sequences outside the proposed mobile locus. As was observed for qacC, the gene can be present on long conjugative plasmids as well as RC-plasmids. In the latter case, the only other gene usually present is rep2. It has been observed before that lnuA in such plasmids is more strongly conserved than rep (Lüthje et al., 2007). Comparison of nine RC-plasmids by multiple nt alignment identified the segment containing DSO, lnuA and SSO (694 bp) as 92.7% conserved, compared to 77% for the rep gene (coding sequences only) or 74.4% for the plasmid excluding the DSO-lnuA-SSO segment (ignoring a weakly conserved region). The conserved flank upstream of lnuA is 78 nt long and the downstream flank is for 160 nt conserved, though the number of direct repeats immediately downstream of lnuA varies. The circular homology plot is available as Figure S2.

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.