The minimal replicon and the replication origin
By constructing deletion mutants we could show that the minimal replicon of pRN1 requires the nucleotides 2101–5197 and that in addition the region between the replication protein gene and the stem–loop is dispensable (pCdel24_25, Fig. 4). These results are in agreement to the results of a recently published study (Joshua et al. 2013) that showed that the region from position 2101 and 5270 is required. In this study as well as in our study, mutations were generated to change the structure of the stem–loop. Deletions of 18–20 bases of the loop region and deletions of only three bases within the loop yielded plasmids which are no longer able to replicate. Our results are consistent with these experiments as complete deletion of the loop as well as other mutations changing the structure of the stem–loop are not tolerated. Thus, both studies agree that the integrity of the stem–loop is required for successful replication of the plasmid strongly suggesting that the stem–loop harbors the replication origin. In addition, both reports agree that mutations changing the GTG motif in the stem–loop do not abrogate the replication capability of the plasmid.
Nevertheless, both studies come to completely different conclusions concerning how the plasmid is replicated. Joshua et al. (2013) suggest that the stem–loop is the double-stranded origin of rolling circle replication and that the stem–loop directly after the stop codon of the replication protein is the single-stranded origin of rolling circle replication. On the contrary, we suggest that pRN1 is replicated in a different way which appears to be conserved in some bacteriophages (see below) and that the stem–loop directly after the stop codon is a transcriptional terminator. In fact we could show that this part of the plasmid can be deleted.
Comparison with other origins of replication
Three different modes of replication have been described for plasmids: the rolling circle replicating plasmids, theta replicating plasmids, and a mode of replication with strand displacement. These modes are carried out by different sets of proteins either encoded by the plasmid or by the host genome. In many cases, the plasmid-encoded proteins have an important role in plasmid replication initiation, whereas the further processive replication is carried out by the host proteins. Therefore, DNA polymerases and other components of the replication fork are rarely encoded on a plasmid. In contrast, bacteriophages that are able to produce a large progeny in short time tend to not rely on the host replication proteins but may also encode the respective proteins by themselves. An overview of the different phage replication modules has been reviewed by Weigel and Seitz (2006).
Previously, it had been proposed that pRN1 replicates by a rolling circle mechanism based on the similarity of the above described stem–loop structure to the origin structure of the rolling circle replicating plasmid pLS1 from Streptococcus agalacticae (Kletzin et al. 1999). However, this mode of replication is not likely to be used by pRN1, as the characterization of the pRN1 replication protein ORF904 did not reveal any endonuclease activity which would be required for this type of replication initiation.
The activities carried out by the pRN1 replication protein might give a hint how the replication is realized. The pRN1 replication protein has a robust and site-specific primase activity, a DNA polymerase activity devoid of proof reading, and a weak DNA unwinding activity with a translocation direction 3′–5′ on single-stranded DNA and high ATPase activity in presence of double-stranded DNA (Lipps et al. 2003; Sanchez et al. 2009; Beck et al. 2010). As reasoned above, it is more likely that the plasmid-encoded proteins participate in replication initiation than in processive replication. Thus, we suggest that the primase–helicase activity of pRN1 is functionally unrelated to the primase–helicase proteins of bacteriophages, for example, gp4 from T7 (Please refer to Table 1 for a comparison of the replication proteins of some model replicons.). In T7 replication the primase–helicase is part of the replication fork and the helicase is encircling the single-stranded DNA lagging strand in 5′–3′ direction helping to unwind the phage genome and the primase is synthesizing with low sequence specificity the primers for Okazaki fragment synthesis (Matson et al. 1983). The same type of activity would be impossible for the pRN1 replication protein as the helicase travels in the opposite direction and is much less processive than the T7 helicase. In fact both helicases are also from different superfamilies: SF3 in case of pRN1 and SF4 in case of T7. Helicases of these two superfamilies assemble as hexameric rings. Especially helicases of superfamily 3 are also able to encircle double-stranded DNA and might also be able to unwind double-stranded DNA at a replication origin, for example, the SV40 protein large T-antigen (James et al. 2004; Hickman and Dyda 2005). In addition, SF3 helicases are structurally similar to ORC proteins which assemble and destabilize the DNA duplex at archaeal and eukaryotic replication origins. Thus, we suggest that the SF3 helicase domain of the pRN1 replication protein is involved in melting the plasmidal replication origin.
Table 1. Comparison of the domain structure of plasmidal and viral replication proteins (see text for details).
In addition, the pRN1 replication protein could function similar to the bacteriophage P4 alpha protein. This multifunctional protein binds with a winged helix DNA binding domain at the repeats present at the P4 replication origin followed by unwinding and priming (Ziegelin and Lanka 1995; Yeo et al. 2002). However, two major differences between the P4 and the pRN1 replication remain. The primase of the P4 alpha protein is of the bacterial type (Topprim domain) and the pRN1 replication origin does not have the iteron structure typical for many plasmids, instead pRN1 has a large stem–loop. Remarkably, the pRN1 replication protein also has a winged helix DNA binding domain at its C-terminus as revealed by careful sequence analysis and confirmed by biochemical data (Sanchez et al. 2009).
Another similar replication initiation is found for the RSF1010 plasmid. In this case, however, three separate polypeptides carry out the replication initiation (Honda et al. 1991; Scherzinger et al. 1991; Miao et al. 1995; Geibel et al. 2009). The RepC protein recognizes the repeats of the iteron replication origin. Next, RepA unwinds the double-stranded DNA exposing short stem–loops (ssiA and ssiB) which are then specifically recognized by RepB' which synthesizes a primer 3′ to the stem–loop. Next host replication proteins extend the primers exclusively in leading-strand mode.
The hallmark of the pRN1 replication protein is its primase activity. We therefore searched the Aclame collection of mobile genetic elements (Leplae et al. 2010) for the presence of primases. We first retrieved from the Conserved Domain database domains with known primase activity and selected one representative protein from each conserved domain. We then queried the Aclame collection using the representative protein, that is, the primase domain of the pRN1 replication protein, the primase domain of the RSF1010 replication protein RepB', and Herpes virus primase as well as the cellular primases of the archeaon Pyrococcus horikoshii (small subunit) and Methanobrevibacter smithii (large subunit), the bacterium Synechococcus elongates and the human Prim/Pol-Protein. Although the Aclame collection contains over 122,000 proteins from 2300 mobile genetic elements, we found only a very limited number of proteins with highly significant (E = 0.01) and borderline significant (E = 1) similarity to primases. We note, however, that proteins of genetic elements may be subject to rapid divergent evolution. Thus, sequence similarity might be lost between homologous proteins. In total, we only found about 50 proteins half of which are related to the bacterial primase DnaG. Thus, it appears that the involvement of primases in plasmid and bacteriophage replication is indeed minor.
Remarkably, however, is the similarity of open reading frames from several linear phages to the primase domain of the replication protein pRN1. A more detailed analysis reveals that the replication proteins from the nearly identical phages BIP-1/BPP-1/BMP-1 (Bordetella bronchiseptica) are highly related to the pRN1 replication protein. The phage replication proteins have nearly the same domain structure as only the domain pRN1_helical is swapped against the PriCT_2 domain (Table 1). The latter domain is also predicted to be helical (Iyer et al. 2005). We analyzed more closely the BIP-1 sequence. With MEME (Bailey et al. 2006), we searched for repeats and found two different repeats with five instances each around nucleotide position 9200 (within a gene coding for a crystalline protein) and another three instances of a 47 bp repeat at position 37,000. Both iteron structures do not appear to be a replication origin. Instead we found 3′ to the replication proteins a large stem–loop as also observed for pRN1 (Fig. 6). We therefore suggest that these types of replication proteins carrying an archaeoeukaryal primase domain, SF3 helicase, and a winged-helix DNA binding domain recognize and assemble at a stem–loop structure.
Figure 6. Mfold predicted stem–loop downstream of the BIP-1 replication protein. The replication protein is on the complementary strand and ends at position 27,177 (in the loop of the small stem–loop at the 3′ end).
Download figure to PowerPoint
We suggest that the replication initiation proceeds in these following steps. Initial binding of the monomeric replication protein downstream of its own gene. Then, assembly of a multimeric complex aided by DNA which could fold into an alternative structure at the stem–loop. The replication protein is known to assemble into hexameric rings in the presence of double-stranded DNA and the nonhydrolysable ATP analog AMP-PnP (Sanchez et al. 2009). ATP hydrolysis by the superfamily 3 helicase domain of the replication protein could power DNA unwinding and finally the unwound single-stranded DNA is used as template for the primase (Fig. 7). It is possible that the helicase once assembled in the two hexameric rings could further translocate and unwind additional DNA stretches. This would explain why the GTG motif within the stem–loop is not required for replication (see above). Further on the host replication proteins come into play. The host proteins might recognize the replication bubble with the primer/templates and build up two replication forks which progresses in both directions. The bidirectional movement suggests itself for symmetry reasons and is more suitable to allow complete replication of the linear BIP-1 phage genome. In phage BIP-1, the putative replication is located at about position 27,000, thus roughly in the middle of the linear genome of 42,638 bp.
Figure 7. Model of replication initiation involving stem–loops. The replication protein ORF904 binds sequence specifically at not yet identified sequence motifs present downstream of its replication origin. Initial DNA contacts are mediated through the winged-helix DNA binding domain, the beta-hairpins of the helicase domain, and eventually also by the prim/pol domain. Oligomerization occurs and is facilitated by the stem–loop structure and powered by ATP hydrolysis. Primer synthesis may occur at the exposed single-stranded DNA of the stem–loop. Alternatively the helicases “pump” double-stranded DNA widening the loop.
Download figure to PowerPoint
During the preparation of the manuscript Joshua et al. (2013) also reported on the delineation of the minimal replicon of pRN1. The group argues that pRN1 is replicated via the rolling circle replication but give no insight how such a replication could be carried out by the replication protein encoded on the plasmid. Rolling circle replication requires an endonuclease which cuts site-specifically within the double-stranded origin (Khan 2000). The replication protein does not have sequence similarity to other rolling circle replication proteins nor to nucleases. Rolling circle replication further requires a single-stranded origin. The sso suggested by Joshua et al. is probably a terminator stem–loop and our experiments show that this structure can be deleted. Thus, in our view a rolling circle replication of the pRN1 plasmid is very unlikely.
Here we demonstrate that the stem–loop structure is a conserved feature within the pRN plasmid family and that similar replication modes might also operate in linear phages. Although we cannot present direct biochemical evidence how the replication proceeds we suggest a replication mechanism which is consistent with the known enzymatic properties of the replication protein.
In summary, we find that pRN1 replication is remarkable in several aspects. Only a single multifunctional replication protein appears to be required to initiate plasmidal replication, replication initiation might take place at a large stem–loop structure and surprisingly the same type of replication might be realized by some linear bacteriophages.