A Brief History of Plasmids

ABSTRACT

In the late 1950s, a number of laboratories took up the study of plasmids once the discovery was made that extrachromosomal antibiotic resistance (R) factors are the responsible agents for the transmissibility of multiple antibiotic resistance among the enterobacteria. The use of incompatibility for the classification of plasmids is now widespread. It seems clear now on the basis of the limited studies to date that the number of incompatibility groups of plasmids will likely be extremely large when one includes plasmids obtained from bacteria that are normal inhabitants of poorly studied natural environments. The presence of both linear chromosomes and linear plasmids is now established for several Streptomyces species. One of the more fascinating developments in plasmid biology was the discovery of linear plasmids in the 1980s. A remarkable feature of the Ti plasmids of Agrobacterium tumefaciens is the presence of two DNA transfer systems. A definitive demonstration that plasmids consisted of duplex DNA came from interspecies conjugal transfer of plasmids followed by separation of plasmid DNA from chromosomal DNA by equilibrium buoyant density centrifugation. The formation of channels for DNA movement and the actual steps involved in DNA transport offer many opportunities for the discovery of proteins with novel activities and for establishing fundamentally new concepts of macromolecular interactions between DNA and specific proteins, membranes, and the peptidoglycan matrix.

INTRODUCTION

In 1887 Robert Koch published the results of several experiments demonstrating that the causative agent of anthrax was the rod-shaped bacterium Bacillus anthracis (1). Approximately 100 years later it was established that this bacterium harbored two plasmids that were required for its virulence properties (2). Genetic evidence for the existence of plasmids initially came from the incredibly insightful studies carried out in the laboratories of J. Lederberg and W. Hayes in the early 1950s (3). These studies, which identified the sex factor, F, as the transmissible agent responsible for the donor state of a conjugative Escherichia coli strain, were built on the earlier groundbreaking published report of genetic recombination in bacteria by J. Lederberg and E. L. Tatum in 1946 (4). It was J. Lederberg who proposed the term plasmid in 1952 for extranuclear structures that are able to reproduce in an autonomous state (5). These were, of course, extraordinary times for the development of the field of molecular genetics since in the early 1950s. J. D. Watson and F. H. C Crick, using chemical and X-ray diffraction data, proposed the double-helix structure of DNA (6). It was only several years later that A. Kornberg and colleagues described the synthesis in vitro of DNA (7).

In the late 1950s, a number of laboratories took up the study of plasmids once the discovery was made that extrachromosomal antibiotic resistance (R) factors are the responsible agents for the transmissibility of multiple antibiotic resistance among the enterobacterial The earliest studies on R plasmids, carried out mainly by scientists in Japan (8), and the pioneering work on colicinogenic (Col) plasmids by P. Fredericq of Belgium in the mid-1950s (9) resulted in the identification of a variety of naturally occurring plasmids that, subsequently, were analyzed in the 1960s for the plasmid properties of autonomous replication, mobility, incompatibility, and host range. The list of plasmids available for study during this period was augmented significantly by the isolation by several laboratories of F-prime factors from E. coli strains that carried a chromosomally integrated form of the F factor (10). The detailed genetic analysis of these F-prime plasmids contributed greatly to our understanding of the chromosomal state of a plasmid and the mechanics of chromosomal gene transfer from donor to recipient bacterium. The replication and conjugal transfer properties of the F-prime factor, Flac, analyzed by F. Jacob and F. Cuzin (11), also led to the development of the concept of plasmid replicons by F. Jacob, S. Brenner, and F. Cuzin in 1963 (11), a concept that greatly influenced approaches to the study of the control of plasmid DNA replication.

To provide a detailed account of all that has transpired from the early 1950s to the present day that has led to our present understanding of plasmid properties is obviously beyond the scope of this limited review. Furthermore, to minimize the number of references for the many key studies in this highly selective and personal rendition of plasmid history, many of the references provided are general review articles or a review of a body of work from a particular laboratory. At the outset I extend my apologies to work that has been slighted, omitted, or, perhaps in the view of my senior colleagues, given too much emphasis.

THE GENETICS ERA

The foundations of plasmid biology were built largely on careful genetic analysis of the properties of the F plasmid, F-prime derivatives, R plasmids, and Col plasmids of E. coli and related enterobacterial and penicillinase (Pnase) plasmids of Staphylococcus aureus. Soon after it was established that the F plasmid could assume both an autonomous and integrated state, F. Jacob and E. Wollman (10) drew analogies between this E. coli sex factor and the temperate phage lambda in coining the term episome. In its strictest interpretation, an episome was considered to be a genetic element that is transmissible and exists in two mutually exclusive states: extranuclear and integrated as part of the bacterial chromosome. Many laboratories adopted this designation for their plasmid under study, often loosely and either with no evidence for a stable chromosomal integrated state or with little consideration of the frequency of insertion of the plasmid into the chromosome. After spirited debate over the use of this term in the late 1960s (12), it was finally decided that the distinction between a plasmid and an episome is for the most part artificial and use of the designation episome was eventually abandoned.

The reports of transfer of multiple antibiotic resistance between E. coli and related bacterial species in the 1950s and early 1960s by several laboratories in Japan and the demonstration that R plasmids are extrachromosomal in nature (8) drew considerable attention from the international community of bacterial geneticists and resulted in a number of laboratories in Europe and the United States taking on the task of isolating and studying the genetic properties of R plasmids (13). Scientists in Japan established early on that transmission of multiple drug resistance required cell-to-cell contact and that R plasmids exhibited many conjugative properties in common with the E. coli F sex plasmid and F-prime plasmids (8). T. Watanabe, a pioneer in the discovery and study of R plasmids, coined the term RTF for the region of the R plasmid responsible for replication and transfer and in a seminal review article presented evidence in support of the notion that R factors contained two genetically distinguishable components, RTF and an antibiotic resistance (r) determinant (14). The demonstration that these two determinants could be separated by transduction provided critical support for this model (15, 16). The separability of these two R determinants suggested that R plasmids were analogous to F-prime plasmids and that possibly multiple-resistance R plasmids arose by stepwise, multiple crossovers at different antibiotic resistance sites on bacterial chromosomes during the passage of the R plasmid through different bacterial strains. In view of our unawareness of the existence of antibiotic resistance transposons at the time, it certainly was not an implausible proposal.

In addition to examining plasmid behavior during intrageneric crosses between different strains of E. coli, a number of R plasmid intergeneric crosses were carried out involving E. coli and other members of the Enterobacteriaceae family. It was noted by several laboratories that the transfer of an R plasmid from E. coli to Salmonella enterica serovar Typhimurium resulted in segregant R plasmids that had lost one or more of their drug resistance markers. Some of the early work by E. S. Anderson with these intergeneric crosses (17) led to the proposal that, in addition to composite R plasmids containing RTF and r determinants, there existed relatively small size, autonomously replicating R plasmids that were not self-transmissible but whose transfer could be promoted by a self-transmissible R plasmid. These R plasmid conjugal transfer studies (17, 18), along with early work on Col plasmids (19), led to a clear distinction between sex plasmids and nonconjugative but mobilizable plasmids.

During the mid to late 1960s, a number of studies were published that dealt with the dissociation of R plasmids into physically separable RTF and r determinants after transfer of a composite R plasmid from E. coli to S. enterica serovar Typhimurium or Proteus mirabilis. In both of these Gram-negative recipient bacteria, certain composite R plasmids that were relatively stable in E. coli showed a propensity to dissociate into RTF and r components in these related bacteria. The phenomenon of dissociation of a composite R plasmid into its two component parts in P. mirabilis became the subject of intense research by various laboratories. Several pioneers in the study of R plasmid structure, including R. Clowes, S. Cohen, S. Falkow, and R. Rownd, engaged in lively discussions on a number of issues, including the mechanism of dissociation of composite R plasmids and the reassociation of their components, whether the smaller r component is self-replicating, the mechanism of formation of multiple forms of the r component, and the mechanism responsible for the increase in r component copy number after extended growth of the bacterium in the presence of a selective antibiotic (20–23). During this time the analysis of the structure of the RTF and r determinants was greatly facilitated by taking advantage of the observed density difference between these two components in cesium chloride (CsCl)-buoyant density gradients and at a later time using the dye-CsCl method to isolate in pure form the RTF and r segments as covalently closed circles whose sizes could be determined by electron microscopy. The R plasmids under study, including R222/R100 (also designated NR1), Rl, and R6, exhibited considerable homology with each other and with the F sex factor. It was not until the early 1970s, when the powerful technique of electron microscopy analysis of heteroduplexes formed from individual DNA strands of duplex DNA molecules was developed (24, 25), that both the overall structure and the mechanism of reversible dissociation of composite R plasmids were determined (26). With this heteroduplexing method, the presence of IS I elements at the boundaries of the RTF and r determinants in the R6 plasmid derivative, R6-5, was demonstrated (27). The identification of insertion sequence (IS) elements at these positions provided a likely mechanism for both the reversible dissociation of composite plasmids and the amplification of the r determinant that were shown to occur in S. enterica serovar Typhimurium and P. mirabilis. The development of the heteroduplexing technique and the identification of IS elements in the early 1970s were critical breakthroughs not only in unraveling at a mechanistic level the reversible dissociation of composite plasmids but also in providing critical insight into the evolution of R plasmids, particularly with regard to the acquisition of single or multiple antibiotic resistance genes.

Largely through the pioneering efforts of Pierre Fredericq in the late 1940s through the 1960s, it was established that Col factors exhibit genetic properties very similar to those of R factors (9). Of the wide variety of Col factors that were examined in E. coli studies with ColEl, ColV, Coll, and ColB, particularly, played a critical role in establishing early on that Col factors were extrachromosomal elements capable of autonomous replication and that they could be categorized into factors that were either conjugative, like the F factor, or not capable of promoting conjugation but transferable as an independent plasmid when a conjugative plasmid was present (9, 19). These early observations with Col plasmids contributed significantly to the concept of nonconjugative but mobilizable plasmids. The finding of derivatives of the ColB and ColV plasmids that contained E. coli chromosomal genes indicated that these plasmids also could integrate with the chromosome, albeit unstably (9). This evidence for an integrated state of certain Col plasmids along with their other F-like characteristics supported the early classification of Col plasmids as episomes along with the F sex factor and the temperate lambda bacteriophage (10).

The medical importance of antibiotic resistance was clearly a major driving force from the beginning in support of basic genetic research on R plasmids. The presence of antibiotic resistance genes on R plasmids, in turn, greatly facilitated the selection of exconjugates from conjugal transfer events between a variety of bacterial species. Given the large number of R plasmid collections obtained from all parts of the globe, attempts were made early on to come up with a classification scheme for various plasmids. One of the first such schemes involved testing whether a plasmid inhibited F-mediated conjugation when both plasmids were present in the same E. coli cell line. On this basis, plasmids, initially R and Col plasmids, were designated as fi + or fi − depending on the ability of the plasmid to inhibit F fertility (28). This inhibition was correctly deduced to reflect similar conjugal transfer systems, including pili and regulatory factors, between the fi + plasmids and the F sex factor. R. Hedges, E. Meynell, and N. Datta played a particularly significant role in critically examining this scheme of classification of plasmids (29, 30). In early studies approximately 50% of the fi + R plasmids were found to determine a pilus that closely resembled the F sex pilus on the basis of sensitivity to sex pilus-specific phage and serological properties. Many, but not all, of the remaining plasmids examined (fi −) specified a pilus with characteristics of the Coll sex pilus. Although initially using fertility inhibition brought some order to the wide array of plasmids under study, it soon became clear that its usefulness as a classification scheme was limited. Not only did it not include nonconjugative plasmids, but it also failed to deal adequately with derepressed mutants of a particular sex factor or, as subsequently found, the wide range of morphologically distinct sex pili. Finally, although information on conjugative plasmids in Gram-positive bacteria was largely missing at the time, it was clear that this classification scheme would not be applicable to plasmids of Gram-positive bacteria. It was, therefore, not surprising that the fertility inhibition test for classification of plasmids was abandoned in the late 1960s and replaced by incompatibility grouping in the early 1970s.

It was found that members of the fi + class of plasmids, while able to coexist with the F plasmid in E. coli and with members of the fi − group, often failed to coexist with all of the other members of their class so that within each group specific pairs of plasmids were incompatible (29), The idea of using incompatibility as the basis of a classification scheme won wide acceptance for plasmids in Gram-negative bacteria through the efforts of several groups, including those of Datta, Hedges, and Meynell (29), and in Gram-positive bacteria through the pioneering studies of R. Novick and M. Richmond on Pnase plasmids of S. aureus (31, 32). Novick followed up on observations in the early 1950s that described the relatively high frequency of loss of the ability of staphylococci to produce Pnase by providing genetic evidence for the plasmid nature of Pnase genes in S. aureus (33). In the absence of a conjugal transfer system in S. aureus, transduction was used to bring about various pairwise combinations in a single cell to test for compatibility using different selective markers for each plasmid. On this basis a number of incompatibility groups were established, and subsequent work by Novick on the Pnase plasmids provided important insight into the mechanism(s) of incompatibility (34).

The use of incompatibility for the classification of plasmids is now widespread. Its use for plasmids of Gram-negative bacteria was greatly advanced by the work of M. Couturier, W. Maas, and collaborators in the 1980s who critically examined the criteria for classifying a plasmid in a specific incompatibility group, including the use of DNA hybridization probes (35). Although there are a number of difficulties in creating pairwise combinations of plasmids from bacteria in their natural environment, this classification scheme has also proven valuable in environmental studies involving the distribution of plasmids, especially when complemented by the utilization of hybridization probes representing different compatibility groups. It seems clear now on the basis of the limited studies to date that the number of incompatibility groups of plasmids will likely be extremely large when one includes plasmids obtained from bacteria that are normal inhabitants of poorly studied natural environments (36, 37).

It was suggested early on that plasmid incompatibility is a consequence of two plasmids sharing common elements responsible for plasmid maintenance, namely, replication control and/or partitioning systems (34, 38, 39). Some of the most definitive demonstrations of the key role of common regulatory elements of replication in the determination of incompatibility came from the groundbreaking in vitro studies of J. Tomizawa (40) and the in vivo work of G. Cesareni (41) and their collaborators on plasmid ColEl. The in vivo studies definitively showed that a single nucleotide base change in the regulatory RNA1 molecule (and consequently a concomitant change in the primer RNAII) results in the creation of a new incompatibility group. In general, however, naturally occurring plasmids that are compatible share relatively few nucleotide sequences in common, whereas different plasmids of the same incompatibility group share a high proportion of nucleotide sequences in common. In addition, it has been observed that conjugative plasmids of the same incompatibility group have closely related or identical conjugal transfer genes and sex pili whereas conjugative plasmids of different groups generally differ in their transfer region and exhibit unrelated pili (42, 43).

Although I have up to this point dealt largely with the contributions of genetic studies on F, R, and Col plasmids, to our understanding of the basic properties of plasmids, it should be emphasized that research on a number of other plasmids in both Gram-negative and Gram-positive bacteria in the early 1970s and the years following also provided novel insights into plasmid behavior. This list includes a variety of hemolysin plasmids in E. coli, streptococcal plasmids carrying antibiotic resistance and fermentative genes; plasmids encoding enterotoxins, adherence factors, or iron-sequestering factors; plasmids of both Gram-positive and Gram-negative bacteria that carry genes determining resistance to toxic inorganic cations; gonococcal antibiotic resistance plasmids; degradative plasmids of Gram-negative bacteria; Ti plasmids of Agrobacterium tumefaciens; plasmids of Rhizobium that are involved in nodulation and nitrogen fixation; and plasmids of various Streptomyces species. Studies on this rich array of plasmids have not only revealed genes and mechanisms responsible for a number of bacterial phenotypic properties of medical, agricultural, or environmental importance, but they have also contributed early on to our fundamental understanding of plasmid replication, mobility, gene acquisition and loss, and signaling processes.

CIRCULAR PLASMID DNA

Both genetic evidence (10) and the autoradiographic analysis of replicating chromosomes (44) in the early 1960s established the circular structure of the E. coli chromosome. The first physical evidence for circularity in DNA came from biophysical studies by W. Fiers and R. Sinsheimer in 1962 on the structure of the single-stranded DNA molecule of bacteriophage phiX174 (45). This was later confirmed by electron microscopy analysis (46). At about the same time the pioneering work by the laboratories of R. Sinsheimer, R. Dulbecco, and J. Vinograd demonstrated the covalently closed, circular form of duplex DNA molecules for polyomavirus particles and for the intracellular RF form of phiX174 bacteriophage (47). The first physical demonstration of the covalently closed form of a plasmid came from electron microscopy analysis of purified ColEl DNA in 1967 (48). A year later J. Vinograd’s laboratory introduced the use of the intercalating dye ethidium bromide to separate covalently closed circular duplex DNA from linear or nicked open circular DNA by equilibrium CsCl centrifugation (49). This procedure greatly simplified the isolation of circular duplex DNA from bacteria, and its use led not only to the identification of a large number of plasmids from a wide array of bacteria but also has provided the means to obtain relatively large quantities of a specific covalently closed circular plasmid element for in vitro analysis.

Well in advance of physical evidence for the circularity of plasmids, extensive genetic analyses of conjugal transfer involving Hfr strains in E. coli and transduction analysis in both E. coli and S. aureus clearly indicated a circular structure for F, Col, and R plasmids (9, 10, 14, 32). The earliest indication of the DNA nature and the size of F and Col plasmids came from experiments involving the radioactive labeling of plasmid elements in vivo with ³²P followed by a kinetic analysis of inactivation (loss) of the plasmid due to ³²P decay (3). It was further shown that the incorporation of ³²P was inhibited by mitomycin C, an inhibitor of DNA synthesis. Although this method was imprecise and frequently overestimated the size of a plasmid element, it provided early support for what was generally assumed to be the DNA nature of plasmids. A definitive demonstration that plasmids consisted of duplex DNA came from interspecies conjugal transfer of plasmids followed by separation of plasmid DNA from chromosomal DNA by equilibrium buoyant density centrifugation (49). Although somewhat laborious, this approach, which preceded the dye-buoyant density equilibrium centrifugation procedure, provided the first effective methods for separating plasmid DNA from chromosomal DNA. This method was based on the finding in the mid-1950s that the base composition of DNA of organisms often showed variation in their mean value. Furthermore, in the late 1950s the P. Doty and M. Meselson laboratories showed that the distribution of base composition of DNA molecules of a specific organism was unimodel within a relatively narrow range (50). In the early 1960s, J. Marmur, R. Rownd, S. Falkow, L. Baron, C. Schildkraut, and P. Doty (51) took advantage of the buoyant density difference between the DNA of Flac sex factors (50% GC) and Serratia marcescens DNA (58% GC) to separate sex factor DNA from chromosomal DNA by transfer of the plasmid to its unnatural host followed by analytical CsCl-buoyant density centrifugation. A similar approach was used with P. mirabilis (39% GC) as the recipient (52). The results clearly demonstrated a correlation between the acquisition of the Flac plasmid, shown by genetic analysis, and the acquisition of DNA with a 50% GC content. The S. Falkow laboratory and, at the same time, the laboratory of R. Rownd used the same method to identify a DNA species corresponding to R factors after transfer of these plasmids from E. coli to P. mirabilis (53, 54). As noted before, the conjugal transfer of certain R plasmids from their natural host, E. coli, to P. mirabilis often resulted in instability of the plasmid, characterized by the reversible dissociation of these plasmids into their RTF and r determinants.

The technique of establishing plasmid DNA in a bacterial host with a significantly different chromosomal GC content was used both to physically identify a plasmid element and to purify that plasmid element. Electron microscopy analysis in 1967 of the DNA corresponding to the F-prime factor, FColVColBtrycys, separated from P. mirabilis chromosomal DNA by buoyant density centrifugation, provided physical evidence for the circularity of a sex factor (55). At about the same time D. Freifelder demonstrated a circular DNA form for the classic Flac plasmid (56). The use of a cleared lysate procedure that removed the bulk of chromosomal DNA from plasmid-containing E. coli cells, developed earlier by the Sinsheimer laboratory for the isolation of phiX174 DNA (57), in conjunction with dye-buoyant density centrifugation, further simplified the isolation and characterization of covalently closed circular plasmid DNA (58). It soon became evident that circularity is a common feature not only of a variety of transmissible F, Col, and R plasmids but also of nonconjugative and mobilizable Col and R plasmids as well as of the lysogenic state of bacteriophage PI (59). In addition, multiple circular DNA forms of a plasmid element often were encountered, particularly if the plasmid was established in a foreign bacterial host (60).

It took over a decade before the notion that all bacterial plasmids were circular DNA elements was dispelled (and even a longer period to correct the notion that bacterial chromosomes were invariably circular). The work of the laboratory of K. Sakaguchi, reported in 1979 and the early 1980s, established that plasmids harbored by the lankamycin- and lankacidin-producing bacterium, Streptomyces rochei, are linear (61). Linear plasmids are now known to occur in a number of different actinomycetes as well as in rhodococci and mycobacteria (62). The presence of both linear chromosomes and linear plasmids is now established for several Streptomyces species (62). The finding of a linear chromosome and 21 linear and circular plasmids in a strain of the spirochete Borrelia burgdorferi (63) is a particularly striking example of the risk involved in making generalizations from the analysis of plasmid or chromosome structures in one or a selected few well-known bacterial species.

PLASMIDS AS GENE CLONING VECTORS: A BOLD ADVANCE

The development of recombinant DNA technology in the 1970s has been the subject of scores of articles and books and, therefore, my treatment of the beginnings of this technology is highly selective. There were three main developmental phases of this technology. The biochemical phase included the demonstration that complementary ends of DNA molecules could be constructed, linked by the enzyme T4 ligase, and introduced into bacteria. The contributions of the laboratories of P. Berg, D. Kaiser, and H. Khorana were particularly critical in this phase. During the same period, the restriction endonuclease EcoRI was isolated and shown to cut DNA at a specific sequence and produce cohesive ends. The laboratories of H. Boyer, J. Mertz, R. Davis, and V. Sgaramella played pivotal roles at this stage. The second developmental phase, which I designate, albeit arbitrarily, the molecular genetic phase, took off with the findings that E. coli cells could be stably transformed with purified antibiotic resistance plasmids using the power of antibiotic selection and that a “foreign” DNA fragment could be inserted into a plasmid in vitro, utilizing the cohesive end generating EcoRl endonuclease followed by linkage of the DNA fragments with T 4 ligase (64, 65). This bold experiment, a collaborative effort involving the laboratories of S. Cohen and P. Boyer, established the great utility of plasmids for gene cloning. The third phase was the development of a variety of plasmid vectors that facilitated gene cloning in bacteria other than E. coli, including Gram-positive bacteria and Gram-negative bacteria distantly related to E. coli. This phase also includes the development of plasmid vectors for use in A. tumefaciens for the generation of transgenic plants.

The storied meeting of S. Cohen and P. Boyer in 1972 in a delicatessen in Honolulu, where their collaboration was discussed and agreed upon, is now part of folklore. A little-known fact is the role of Tsutomu Watanabe in providing the venue for their meeting. Watanabe contacted me in 1970 to determine interest in organizing a joint U.S. and Japan conference on plasmids as part of a National Science Foundation Program for the support of U.S.–Japan scientific exchange on a specific scientific topic. Such meetings were limited to a total of 30 scientists. Stan Cohen joined us as a co-organizer for the U.S. side. We learned in September 1972 that Watanabe was unable to attend the meeting because of health problems and that M. Tomoeda would replace him as a co-organizer. Tragically, Watanabe died in November 1972, less than 2 weeks before the start of the meeting at the East-West Center in Honolulu. Because of his detailed preparations for the meeting, the conference took place as scheduled, providing a wonderful opportunity for many of the U.S. scientists working on molecular aspects of plasmid biology to interact with some of the key scientists of Japan who made significant contributions to our early understanding of plasmid properties. The Japanese contingent included T. Arai, S. Hiraga, K. Matsubara, T. Miki, S. Mitsuhashi, K. Mizobuchi, J. Uchida, H. Hashimoto, and, of course, M. Tomoeda.

Three key initial events in the success of the Cohen-Boyer collaboration were (i) establishing stable plasmid transformants in E. coli and the use of antibiotics for the selection of these transformants; (ii) the use of a restriction endonuclease (EcoRI) that cleaved a plasmid vector (pSC101) at a single site; and (iii) the combination of restriction endonuclease-cleaved DNA fragments with a plasmid vector cleaved with the same restriction enzyme (EcoRI) followed by stable establishment of the recombinant molecule as a plasmid in an E. coli cell (64). The fortuitous presence of a single site in plasmid pSC101 that was recognized by the then-available EcoRI restriction endonuclease, which produced cohesive ends, was certainly a welcomed finding that led to a series of publications that left no doubt that plasmids were powerful tools for cloning DNA from virtually any source in bacteria (66). Interestingly, the genes for the EcoRI restriction endonuclease and modification methylase were later shown to be present on a naturally occurring plasmid (pMBl) that is closely related to ColEl (67). It is perhaps ironic that the ColEl plasmid proved to be an important addition early on to the list of plasmid vectors favored for gene cloning, particularly because it possessed a single EcoRI site and it was stably maintained at a high copy number that could be further amplified by inhibition of protein synthesis in the bacterial host by the addition of chloramphenicol (68, 69). The ColEl-related plasmid pMBl in turn was used as a starting plasmid for the construction of the heavily used pBR and pUC families of cloning vectors in E. coli (70, 71).

Although E. coli was and continues to be the preferred bacterium for gene cloning, the need for developing plasmid cloning vectors in a wide range of bacteria of medical, agricultural, and commercial importance was recognized early. Initially, derivatives of the broad-host-range plasmids RK2 (72) and R300B (RSF1010) (73, 74) were constructed and shown to be effective in establishing genes in most Gram-negative bacteria. Similarly, broad-host-range and narrow-host-range plasmids of Gram-positive bacteria were developed for gene cloning in a number of genera including Bacillus, Staphylococcus, Streptomyces, and Streptococcus. The development of electroporation for the introduction of DNA into a variety of Gram-negative and Gram-positive bacteria greatly facilitated the introduction of in vitro generated recombinant DNA molecules into bacteria that otherwise were difficult to transform by natural transformation or chemical transformation techniques (75).

Although the emphasis of this historical review is on plasmids, it is important to note that once the fundamental principles of gene cloning in bacteria were established, a variety of lambda phage derivatives (e.g., the Chiron phage series) were developed as gene cloning vectors for not only introducing “foreign” DNA into bacteria but also for providing a high level of containment of the bacteriophage with its insert to address biosafety concerns (76, 77). In the 1970s, beginning with the Asilomar conference on recombinant DNA in 1975, there was much discussion about the virtues of plasmids versus lambda bacteriophage for the cloning of genes in E. coli. The laboratory of R. Curtiss took a leadership role in the development of E. coli strains that provided a high level of containment of a plasmid construct by introducing mutations that made E. coli dependent on nutrients not normally found in the environment and/or produced increased sensitivity to natural environmental conditions (78). Eventually, the use of plasmid vectors dominated the field of gene cloning for a variety of reasons, including their ease of use by scientists not schooled in bacteriology and increasing confidence in the safety of gene cloning techniques. This section on gene cloning should not be left without pointing out the incredible importance of the development of gene cloning techniques to research that has led to our present understanding of basic plasmid DNA biology, including plasmid replication and conjugal transfer, plasmid transposable elements, and plasmid dynamics during cell growth and division.

REPLICATION CONTROL CIRCUITRY

Both genetic studies on the conjugal transfer properties of F, F-prime, R, and Col plasmids and the use of acridine dyes to “cure” bacteria of a particular plasmid element provided early evidence for the extrachromosomal or autonomous nature of plasmids. The landmark paper of Jacob, Brenner, and Cuzin (11) proposed that a genetic element, such as a chromosome or plasmid (episome), constituted an independent unit of replication, or replicon, that replicates as a whole. As part of this proposal, the authors hypothesized that each replicon controls the synthesis of an initiator, which acts specifically on its operator of replication (replicator), and that this interaction is required for the initiation of replication. The replicon hypothesis greatly influenced thinking among plasmid workers in the mid-1960s. In proposing their model, the authors rejected the notion that replication control of a specific replicon involved a cytoplasmic repressor that acted on a receiver (replicator) analogous to the Jacob and Monod repressor-operator model for the control of gene expression, but instead they favored a system of positive regulation. A role for a cytoplasmic membrane/surface structure of the bacterium was a critical feature of their replicon hypothesis. This structure was thought to activate the initiator-replicator complex at a specific time in the cell division cycle, allowing replication to proceed along the circular plasmid element. The idea of a plasmid attached to specific membrane/surface structures that are limited in number also had the appeal of providing a mechanism not only for the control of plasmid copy number but also for the partitioning of plasmids during cell growth and division and for the high frequency of transfer of a plasmid (F) despite its low copy number. That some of the basic features of the replicon model have since been proven to be incorrect does not diminish the important effect that this model had in constructively guiding the direction of replication studies by plasmid researchers. It was not until several years later that, largely through the arguments of R. Pritchard (79), a model of negative control of plasmid DNA replication grew in favor among workers in the field. In addition, despite the appeal of a model that includes a role for the cytoplasmic membrane in the initiation and/or replication of plasmid and chromosomal DNA and some evidence in favor of such a connection (80), the unequivocal establishment of this role has proved elusive.

In 1968, a symposium at Cold Spring Harbor, New York, brought together a number of investigators studying the replication of DNA in microorganisms. Of particular interest to the plasmid field were the presentations demonstrating that the prophage state of bacteriophage PI consists of a covalently closed DNA form that replicated autonomously (59) and describing a relatively stable plasmid form of a bacteriophage lambda mutant (lambda dv) (81). Plasmids PI and lambda dv would prove to be very important systems in subsequent studies on the regulation of plasmid DNA replication and, in the case of P1, the partitioning of plasmids during cell division. This meeting, however, included very few presentations dealing with plasmid DNA replication, which was in sharp contrast to the 1978 symposium at Cold Spring Harbor on replication and recombination.

The decade between the two conferences was a period of great advances in our understanding of plasmid DNA replication, and this was reflected in the number of papers dealing with the replication of plasmids in Gram-negative bacteria presented at the 1978 meeting. These presentations included both in vivo and in vitro studies on plasmids ColEl, pMBl, F, Rl, R6-5, R6K, and RK2. Interestingly, at the 1968 symposium, W. Gilbert and D. Dressier (82) reviewed the evidence for a rolling-circle mechanism of replication of bacteriophage DNA, including phiX174 and lambda, and the replicative transfer of DNA during conjugation. In this same paper they presented arguments in favor of a rolling-circle (RC) mechanism as a general mechanism for the replication of plasmid DNA molecules. It was not, however, until much later that the RC mechanism of replication of plasmids was shown to be the predominant mode in Gram-positive bacteria. In contrast and with few exceptions, the theta mode of replication, involving covalently closed circular DNA intermediates, was found to be the norm for plasmids naturally found in Gram-negative bacteria.

In 1971, R. Clowes coined the terms stringent for R plasmids that were maintained at one to two copies per chromosome and relaxed for R plasmids that were normally maintained at many copies per chromosome (20). The so-called relaxed plasmids, e.g., ColEl, generally were found to be much smaller in size than the low-copy-number plasmids. At that time it was also known from density-shift experiments that for both low- and high-copy-number plasmids, the copies are selected randomly for replication (83–85). This observation had significant impact on considerations of various models for the regulation of plasmid copy number. Other important findings in this pre-recombinant DNA era of plasmid research included the isolation of plasmids from Gram-negative bacteria in the form of a DNA-protein complex (designated relaxation complexes), the determination that the replication of the ColEl plasmid in vivo did not require newly synthesized protein, and the presence of ribonucleotides in intact supercoiled ColEl plasmid under conditions of replication in E. coli in the absence of protein synthesis (86). It was also shown that the ColEl plasmid required DNA polymerase I for replication, and this finding subsequently proved to be useful for the screening of replication-defective mutant plasmids because a cointegrate consisting of a ColEl derivative and the plasmid under study was capable of replication in a wild-type E. coli strain but not in a pol I-defective mutant strain (87). During this same period, as discussed earlier, the laboratories of R. Clowes, S. Cohen, S. Falkow, and R. Rownd established the composite nature of several naturally occurring R plasmids and the ability of the RTF and r determinant components to reversibly dissociate into autonomous covalently closed forms (20–23). It was further shown that the frequency of dissociation depended on the bacterial host and that the copy number of each component varied with the growth condition.

The advent of in vitro recombinant DNA technology in the early 1970s gave rise in the following years to a burst of information on plasmid DNA structure, including defining the essential elements for the initiation of replication and the circuitry for the control of plasmid copy number. Utilizing specific restriction endonucleases, it was determined that replication of a plasmid begins at a specific site (region) (88), or in the case of plasmids with multiple origins of replication, at two or more specific sites (88, 89). It was further shown that elongation from this origin can be unidirectional (e.g., ColEl and RK2) or bidirectional (e.g., F and R6K). In the case of the R6K plasmid, a specific terminus region was identified that functioned in R6K or when inserted into an unrelated plasmid (90). The subsequent analysis of this terminus region by the laboratory of D. Bastia provided fundamental information on DNA-protein interactions responsible for the delay or termination of DNA replication fork movement (91). For most plasmids, regardless of size, the replication origin, the initiation protein gene, and other controlling elements were found to be clustered enabled the use of a specific restriction endonuclease to isolate a relatively small DNA fragment containing these essential replication elements. The availability of restriction enzyme-derived DNA fragments containing an antibiotic resistance gene led to the isolation of so-called mini-replicons initially from Flac in the mid-1970s (92, 93) and later from a variety of plasmids of Gram-negative and Gram-positive bacteria. Further restriction endonuclease analysis of plasmids demonstrated that, in many but not all cases, the replication initiation protein could act in trans on a replication origin, allowing for the separation of the origin from the initiation gene and the maintenance of the covalently closed form of the origin fragment in a bacterium that also harbored a compatible plasmid carrying the initiation protein gene (88, 94). The availability of these mini-replicons that were derived from relatively large plasmids greatly facilitated the introduction of mutations in key regulatory elements to determine their effect on the copy number of a plasmid. An elegant series of studies, carried out in vivo in E. coli with plasmid R l by the laboratory of K. Nordstrom in the early 1980s (95), firmly established that a negative control mechanism involving small antisense RNA molecules is responsible for R1 copy-number control.

Several years earlier, the laboratory of J. Tomizawa developed an in vitro replication system for plasmid ColEl, and in a stunning set of publications he and his coworkers worked out the details of initiation of plasmid ColEl replication and the key role of a small antisense RNA molecule in the regulation of the frequency of initiation (96). In vivo studies on the regulation of ColEl replication proved consistent with the Tomizawa model of negative replication initiation control of this plasmid (40, 41). Shortly after these in vitro and in vivo studies with plasmid ColEl and the IncFII plasmids, the laboratory of R. Novick provided similar evidence for an antisense control mechanism involving small RNA molecules for the control of replication initiation of plasmid pT181 (34), a member of a large family of plasmids that naturally occur in S. aureus and other Gram-positive bacteria. Through his research and insightful review articles, Novick also made seminal contributions to our understanding of the molecular basis of plasmid incompatibility (34, 38, 39). Intense in vivo and in vitro studies on individual members of several families of RC replicating plasmids, including pT181, pC194, and pMV158, particularly, by the laboratories of R. Novick, S. D, Erlich, S. Kahn, and M. Espinosa, defined the biochemical steps in the replication of these plasmids and the role of small RNA molecules in the regulation of initiation of replication and plasmid copy number (97). Striking features of RC replication that came out of this work were the formation of a covalent linkage between the initiation protein and a specific sequence at the double-stranded origin, the formation of a single-stranded circular DNA intermediate, and the biochemical mechanisms by which small RNA molecules controlled the synthesis of the replication initiation protein. The formation of a single-stranded DNA intermediate became the signature of a plasmid that replicated by the RC mechanism (98).

A remarkable feature of many of the plasmids that replicate by the RC mode is their promiscuity among Gram-positive bacteria (99). This is perhaps different from most plasmids of Gram-negative bacteria, which are limited in their host range. Notable exceptions are the well-studied plasmids RSF1010 and RK2, which have an extended host range among Gram-negative bacteria. Plasmid RK2 is a member of the IncP alpha incompatibility group and was isolated in 1969 at a Birmingham (United Kingdom) hospital (100). The structure and replication properties of RK2 received particular attention soon after in vitro recombinant DNA techniques were introduced. It is a member of a broad class of plasmids that are characterized by the presence of direct nucleotide sequence repeats (iterons) at the replication origin. RK2 encodes two forms (differing in size due to the presence of an additional 98 amino acids at the N terminus of the larger protein) of a replication initiation protein (94). The presence of iterons at a replication origin was first demonstrated for plasmid R6K in 1979 (101). Since that time, a wide range of plasmids, particularly in Gram-negative bacteria, were shown to contain iterons within their origin. This list includes R6K, RK2, F, PI, pSClOl, Rtsl, pPSlO, and lambda dv among the studied plasmids (102). The replication initiation protein for iteron-containing plasmids was found to work in trans and also was shown in some cases to be autoregulatory in the control of its expression (103). In addition, the iteron sequences themselves were found to be strong incompatibility determinants in vivo and, at least for several of these plasmids, the presence of excess levels of replication initiation protein did not increase plasmid copy number or overcome iteron inhibition (102). These observations suggested a regulatory mechanism for the control of copy number that differed from regulation by antisense RNA or by limiting the concentration of the replication initiation protein. A so-called handcuffing or coupling model was proposed in the late 1980s for the control of plasmid copy number of plasmids containing iterons at the replication origin (104, 105). This model drew further support from the finding that mutants of the replication initiation protein that exhibited a copy-up phenotype were defective in vitro in coupling or handcuffing replication origins at their iteron sequences (102, 103).

The replication initiation proteins and origin of replication of plasmid RK2 are active in a wide range of bacteria. More recent studies on this replicon have demonstrated its ability to use different mechanisms of recruitment of host DnaB helicases, depending on the particular host and whether the larger or smaller form of initiation protein is present (106). Considerable work was also done on the coordinately regulated sets of kil (host-lethal or -inhibitory) and kor (Jb7-override) genes on plasmid RK2, particularly by the laboratories of D. Figurski (107) and C. Thomas (108) in the early 1980s. These sets of genes, which undoubtedly contribute to the stable maintenance of plasmid RK2 in a wide range of bacteria, have been shown by the Thomas laboratory to be part of an intricate network that regulates the expression of genes involved in replication, conjugal transfer, and, possibly, plasmid partitioning (109). The complete nucleotide sequence of this complex plasmid was reported in 1994 (100).

The strategy employed by plasmid RSF1010 to extend its host range has been shown to differ from that used by RK2. RSF1010, whose complete sequence was reported in 1989 (110), is a member of the IncQ group of plasmids along with the closely related plasmids R1162 and R300B (111). Particularly through the efforts of E. Scherzinger and M. Bagdasarian and colleagues, RSF1010 was found to encode for three proteins essential for its replication (112, 113). These three proteins, subsequently shown to have helicase, primase, and plasmid replication initiation activities, respectively, enable the plasmid to replicate independently of the host-encoded DnaA, DnaB, DnaC, and DnaG proteins. Thus, in contrast to plasmid RK2, which requires all four of these replication initiation proteins to be encoded by the host for replication in E. coli, RSF1010 clearly has devised the strategy of independence of at least several key host replication proteins for its maintenance in E. coli and distantly related hosts. Remarkably, the range of stable maintenance of RSF1010 has been shown to include several Gram-positive bacteria (114). Undoubtedly, as other plasmids with an extended host range in either Gram-negative or Gram-positive bacteria are studied, the list of strategies to adapt to different host environments will grow.

One of the more fascinating developments in plasmid biology was the discovery of linear plasmids in the 1980s. An initial understanding of the structure of the ends of linear plasmids and their mode of replication was only achieved in the 1990s. The laboratories of S, Cohen and C. W. Chen were particularly instrumental early on in providing critical information on the nature and position of the replication origin of the linear plasmids of Streptomyces and mechanisms of duplication of telomeres as part of the replication process (115, 116). Not unlike the many variations on the several basic themes of replication of theta and RC replicating plasmids found among naturally occurring circular DNA plasmids, it is reasonable to expect that there will be additional novel findings with regard to the stable maintenance of linear plasmids as the number of these plasmids under study increases.

PLASMID MOBILITY

One of the major success stories of plasmid biology is the unraveling of the mechanism of regulation of the F conjugation system. The stepwise progress that was made by a number of investigators over a 20- year period up until the mid-1980s has been chronicled by N. Willetts (117), a leading early investigator of F plasmid mating-pair formation and conjugal transfer. Research contributions by a number of laboratories have progressively led to the identification of an intricate system of operons, genes, and regulatory elements that are responsible for the conjugal transfer properties of this sex factor. These research findings began with the fortuitous use of an E. coli strain carrying a derepressed mutant form of F plasmid by Lederberg and Tatum (4). Their insightful experiments, which led to the discovery of bacterial sexuality, were followed by the equally important work on F plasmid-mediated transfer of chromosomal genes by W. Hayes, E. Wollman, and F. Jacob in the 1950s (3) and extensive complementation analysis of transfer-defective F and R plasmids (117). These genetic contributions provided the basis for the subsequent biochemical analysis of mating-pair formation and conjugal DNA transfer involving several different plasmids in Gram-negative bacteria. Although it is clear that F plasmid is the paradigm for conjugal transfer processes in Gram-negative bacteria, substantial progress in research on plasmid transfer among Gram-positive bacteria has led to our understanding of fundamentally different processes of mating-pair formation and conjugal DNA transfer, including the identification of pheromone-responding conjugative plasmids in the enterococci (118–120).

Although much yet lies in wait for discovery from genetic approaches to plasmid-mediated conjugation, including the mechanism of linear plasmid transfer in the actinomycetes and other bacteria, the present era of bacterial sexuality research is largely focused on biochemical and molecular biology approaches. The biochemical study of plasmid DNA transfer likely began with reports by M. Ohki and J. Tomizawa (121) and W. Rupp and G. Ihler (122) at the Cold Spring Harbor meeting of 1968 describing a clever set of experiments that demonstrated both the specific transfer of one of the two preexisting strands of F plasmid to an E. coli recipient starting from a 5′- end and the replication of the complementary strand that was left in the donor cell. The development of this model of RC replication was influenced by earlier studies of the replication cycle of phiX174 by the laboratory of R. Sinsheimer in the 1960s (123), It has been borne out by subsequent biochemical analyses of plasmid conjugal transfer in both Gram-negative and Gram-positive bacteria. The finding in the late 1960s of naturally occurring ColEl and F plasmids in the form of relaxation complexes consisting of covalently closed circular DNA and proteins that could be induced to carry out a site-specific nick by treatment with certain denaturing agents demonstrated the existence of plasmid DNA-protein complexes that conceivably could carry out the initial nicking event in conjugal DNA transfer (58, 86). Initially, it was considered that these complexes possibly were involved in the initiation of plasmid DNA replication via an RC mechanism and/or the conjugal transfer of the plasmid (86). Evidence was obtained later demonstrating a role of the relaxation complex in conjugal transfer (124, 125). The landmark studies of the Tomizawa laboratory in the 1970s that described biochemical events in the initiation of replication of ColEl strongly argued against RC replication and established a theta mode of replication of this plasmid (96). Studies on the ColEl relaxation complex, however, established both the site specificity of the opening event and the covalent linkage of a protein to the 5′-end of the nicked strand as part of the conjugative transfer process. However, conditions could not be obtained for the separation of the protein(s), presumed to be responsible for the nicking event, from the ColEl complex and its purification in an active form. It was several years later that the laboratory of E. Lanka, working with relaxation complexes of plasmid RP4/RK2, designated relaxasomes, developed conditions for the purification of the enzyme (relaxase) responsible for the site-specific relaxation event plus associated proteins (126, 127), This work represented a critical breakthrough in our understanding at the molecular level of the initial nicking event in conjugal DNA transfer and also provided the means for a structure-function analysis at the in vitro level of an origin of transfer (oriT). The nucleotide sequences of the oriT site of a number of conjugative plasmids have now been defined as well as for this site on mobilizable plasmids, first described by the work of D. Sherratt’s laboratory with ColEl (128, 129) in the 1980s. An especially interesting finding by the laboratory of D. Guiney (130) was the similarity between the oriT regions of the IncP plasmids and the border sequences of Agrobacterium Ti plasmids that serve as nick sites for the transfer of DNA to plant cells. A remarkable feature of the Ti plasmids of A. tumefaciens is the presence of two DNA transfer systems (131, 132). One of the systems, tra, is responsible for the conjugal transfer of the plasmid between bacteria, whereas the second system, vir, is essential for the transfer of the T-DNA within the Ti plasmid to the plant genome. By the early 1990s functional relaxase enzymes were purified from several other plasmids, including members of the IncQ, IncF, and IncW incompatibility groups of Gram-negative bacteria (133). Functional domains of the relaxases have now been defined from mutation studies and a comparison of amino acid sequences based on nucleotide sequence analysis.

Both in vivo and in vitro studies have also defined some of the roles of accessory proteins in the conjugal transfer of DNA. The biochemical and biophysical analysis of a number of tra gene proteins involved in mating-pair formation and DNA transfer represents a relatively new and exciting chapter in research on bacterial conjugation and plasmid transfer. The formation of channels for DNA movement and the actual steps involved in DNA transport offer many opportunities for the discovery of proteins with novel activities and for establishing fundamentally new concepts of macromolecular interactions between DNA and specific proteins, membranes, and the peptidoglycan matrix. This is true not only for intra- and intergeneric transfer of plasmids between bacteria, but also the more exotic DNA transfer process involving the transfer of T-DNA of Ti plasmids from a bacterium to a plant cell. In this latter instance, while there have been great advances in identifying low-molecular-weight compounds involved in bacterium-plant signaling and the initial steps of TDNA transfer, little is known about the process of channel formation between the cells of these two kingdoms or the actual movement of DNA from the bacterium to the plant nucleus.

During the last decade of the 20th century, our understanding of the control of conjugal transfer of plasmid DNA among Gram-positive bacteria also advanced. One notable advance was the purification and characterization in the laboratory of M. Espinosa (134) of a relaxase involved in the mobilization of the Gram-positive RC replicating plasmid, pMV158, with properties consistent with the transfer of the single-stranded DNA form of this plasmid by an RC mechanism. Much remains to be done in characterizing tra proteins of plasmids naturally occurring in Gram-positive bacteria that are involved in mating-pair formation, DNA processing, and DNA transfer. The pioneering work of D. Hopwood in the early 1970s on fertility plasmids in Streptomyces coelicolor (119) was followed much later by the identification and characterization of both linear and circular plasmids in a number of different actinomycetes. Although little is known at this time about the actual events involved in DNA transfer within the actinomycetes, the somewhat special features of this group of bacteria, including the formation of mycelia, promise to unfold fundamentally novel information on DNA movement between cells. As discussed earlier, the analysis of replication of Streptomyces linear plasmids has already provided novel insight into the initiation and termination of replication. A similar analysis of proteins and events in the conjugal transfer of these linear plasmids in the actinomycetes should prove equally interesting.

In the mid- to late 1970s there were a number of reports on conjugative plasmids in Streptococcus/Enterococcus (135). Particularly intriguing was the finding from the laboratory of D. Clewell that certain recipient cells of Enterococcus faecalis produced pheromones that induce cell-to-cell contact or aggregation leading to high-frequency conjugal transfer of a sex plasmid (136). Subsequent studies by the laboratories of D. Clewell and G. Dunny were largely responsible for defining the chemical nature of the pheromone peptides and many of the genes involved in pheromone-induced expression of the conjugative transfer system (118, 136). Continued analysis of the conjugative plasmids is providing a detailed picture of the regulation of expression of transfer genes in a Gram-positive plasmid as well as insight into the increasingly important phenomenon of communication between cells that are physically separated. The extensive studies of E. faecalis plasmids by the Clewell laboratory also resulted in the identification in the early 1980s of a new class of conjugally transferable elements that were named conjugative transposons (135). These elements, which have now been found in a number of other Gram-positive species and in a few Gram-negative bacteria, normally reside as a linear insertion in the bacterial chromosome. During conjugal transfer they assume a circular DNA form that is an intermediate in transposition, Conjugative transposons represent a diverse group of mobile elements that undoubtedly play a major role in the spread of antibiotic resistance. The discovery of conjugative transposons underscores the wide range of conjugative DNA transfer processes in bacteria. The diversity of sexuality in bacteria, at least mechanistically, is truly remarkable, and with the continued selective pressure on bacteria through the use of antibiotics and other selective agents, there is little doubt that additional variations in the mechanism of conjugal transfer of DNA in existing plasmids will evolve. It is also possible that fundamentally different mechanisms of DNA transfer will develop in response to continued selective pressures.

Conjugative transposons are, as we now know, but one of a growing list of mobile genetic elements that includes IS elements, transposons, integrons, and mobile pathogenicity islands. IS elements were initially detected in the 1960s by genetic methods, particularly the finding that a loss of function of a gene that is part of an operon frequently resulted in polar effects on genes located downstream from the mutated gene. One of the earliest such observations is the work of E. Lederberg, published in 1960, that involved spontaneously derived Gal minus mutations (137). Several years later, the laboratory of P. Starlinger characterized similar revertible Gal minus mutants (138), Comparisons of wild-type and mutant bacteriophage lambda gal DNA in the late 1960s using biophysical techniques revealed the presence of extra DNA that was homologous in sequence to more than one Gal minus mutant (138). In the early 1970s, a series of papers were published involving the use of the heteroduplex technique, developed by R. Davis and N. Davidson and by B. Westermoreland, W. Szybalski, and H. Ris (24, 25), that demonstrated by electron microscopy analysis the presence of two distinct IS elements, designated IS1 and IS2, in mutants within the gal and lac operons in bacteriophage lambda. These two IS elements were isolated in the mid-1970s by the Starlinger group (138), and the IS1 element was sequenced by the laboratory of E. Ohtsubo in 1978 (139). During the years that followed several hundred IS elements were identified and many sequenced. They have been grouped into a number of different families on the basis of the activity of their specific transposase and the mechanism of transposition (140). Of special interest was the finding of IS elements on the F sex factor in the mid-1970s by the laboratories of N, Davidson and R. Doenier (141, 142). Their studies and the later work by the Ohtsubo laboratory mapping IS elements in the E. coli chromosome (139) provided strong support at the molecular level for the mechanism of RecA-dependent Hfr formation via the reversible integration by recombination at specific IS sites between the F plasmid and the E. coli chromosome. This mechanism of F insertion proved to be basically the Campbell model of a reversible single and site-specific recombination event between two circular DNA elements, integration of bacteriophage lambda into the E. coli chromosome (143).

A historical treatment of plasmids would be incomplete without some consideration of transposons, which have played an integral part in decades of research on plasmids in addition to the continuing role of insertions and excisions of transposons in bringing about structural changes in plasmids. The number of transposons that have been identified has grown enormously since the initial report of Hedges and Jacob in 1974 on the transposition of ampicillin resistance from the IncP plasmid RP4/RK2 to other replicons (144). Shortly thereafter, the laboratories of M. Richmond, S. Falkow, and S. Cohen made significant contributions to our understanding of the properties of ampicillin transposons, including their mechanism of transposition (27, 145, 146). Around the same time several other laboratories were identifying and characterizing transposons carrying resistance to other antibiotics and to mercury. In the next 2 decades transposons carried by plasmids in both Gram-negative and Gram-positive bacteria were shown to transport a variety of genes besides antibiotic resistance, including catabolic genes, heavy metal resistance, and virulence determinants. Classification schemes were developed for the rapidly increasing number of identifiable transposons, based on the presence or absence of IS elements at their ends (composite or noncomposite), transposition mechanism (replicative or nonreplicative), and sequence similarity of the transposases. Great strides have been made in understanding the mechanisms of transposition involving a complex process of specific breaking and joining of DNA sequences and DNA replication (140). With the discoveries of composite transposons and conjugative transposons, one cannot help but be fascinated not only by the role of mobile elements in shaping or rearranging plasmid and chromosome structure, but also by how one or more specific components of basic mobile elements form new combinations and new classes of elements that are capable of intracellular or intercellular mobility.

STABLE MAINTAINING OF PLASMIDS

It was apparent in the earliest stages of plasmid research that certain plasmids, like F, were stably maintained at a very low copy number without any selective pressure despite the high probability of loss of the plasmid if segregation of copies to daughter cells was a totally random process. The proposal of an active mechanism of plasmid DNA segregation was first made by Jacob, Brenner, and Cuzin in 1963 (11). While an active plasmid partitioning mechanism is most likely the most important mechanism of maintenance of low-copy-number plasmids in bacteria, two other processes, the resolution of plasmid DNA multimers that are generated by recombination or replication events and a fallback mechanism of post-segregational killing (PSK), are now known to also play a role in plasmid stability. The findings by D. Sherratt’s laboratory in the early 1980s of an in cis-acting site, designated cer, in plasmid ColEl and E. coli-specified recombinases that act at this site to resolve ColEl multimers clearly revealed an important mechanism of multimer resolution of a plasmid clement (147). These studies further showed that the relatively small ColEl plasmid relies on host recombinases for the resolution of multimers. In the case of other plasmids in Gram-negative bacteria, e.g., plasmid RK2, and plasmids of Gram-positive bacteria that replicate via a theta-type mechanism, both the resolution site and the resolvase enzyme are encoded by the plasmid element (148, 149). The finding of the lox/cre resolvase system in plasmid P1 not only revealed an important mechanism for the maintenance of this bacteriophage, but this system has also provided an extremely important experimental tool in the genetic modification of the genomes of prokaryotes, plants, and animals (150).

The work of the laboratory of S. Hiraga in the early 1980s not only resulted in the discovery of the par (sop) partitioning region of plasmid F but also identified the presence of two genes, ccdA and ccdB, within an operon that stabilized plasmid maintenance in E. coli by a mechanism that was not replicon specific (151). Initially, it was thought that the stabilization mechanism involved the inhibition of cell division by the CcdB protein when a plasmid carrying the ccd genes was lost or fell below a critical copy number. The role of the CcdA protein was thought to involve suppression of this inhibitory action of CcdB. Further work by the Hiraga laboratory showed that the actual mechanism of stabilization of a plasmid was the killing of the host cell by the ccdB gene product as a consequence of the loss of the plasmid (152). In a series of elegant biochemical studies in the 1990s the Couturier laboratory demonstrated that the CcdB protein binds to the A subunit of E. coli DNA gyrase and inhibits its activity (153). This activity of the CcdB protein results in cell death if the inhibition is not prevented by the presence of the CcdA protein, which forms a tight complex with the CcdB protein. It was further shown that the E. coli Lon protease readily degraded the CcdA protein, thus requiring its continuous syntheses for protection of the host against gyrase poisoning by the relatively stable CcdB protein (154). Active PSK systems have now been described for a number of plasmids in both Gram-negative and Gram-positive bacteria (149). The F plasmid PSK system is representative of a class of proteic systems that consist of low-molecular-weight toxin and antitoxin proteins. This group includes the ParD and ParE proteins of RK2, Kis and Kid of Rl (identical to PemI and PemK of R100), and phd/doc of PL Interestingly, it was found recently that ParE of RK2 also inhibits E. coli gyrase (155) whereas the Kid protein of Rl was shown by the laboratory of R. Diaz to inhibit DnaB-dependent initiation of DNA replication in vitro (156). The elucidation of the mechanism of action of additional PSK toxin proteins is likely to result in the identification of other cellular targets for PSK.

A non-proteic class of PSK systems is best represented by the hok/sok system of plasmid Rl. This system was first discovered in the mid-1980s by the laboratory of S. Molin and then extensively studied by the K. Gerdes group (149). The hok/sok system, which also works on heterologous plasmids, is the best studied of the group of PSK systems that involve an antisense mechanism for the regulation of the low-molecular-weight hok toxic protein. Surprisingly, hok homologous systems have now been identified not only on different plasmids but also on the bacterial chromosome where multiple homologues have been found (149). It is certainly of great interest to determine the functionality, if any, of these hok-like systems located on the bacterial chromosome. A clear role for the RK2 PSK system in the stabilization of this plasmid has been shown by the destabilization effect of a precise deletion of the parDE genes (157). It is also curious that there is now evidence, largely from the relatively recent work of I. Kobayashi, for the idea that a restriction-modification system, known for a number of years to prevent the establishment of viral or unmodified DNA in a bacterium, can also serve as a PSK system (158).

It was not until the early 1980s, through the work of S. Hiraga’s laboratory with F and S. Austin’s laboratory with PI (159), that a genetic region was identified that was likely involved in plasmid partitioning. Shortly after, the work of K. Gerdes and S. Molin (160) and the laboratory of R. Rownd (161) described a similar partitioning (par) region in the plasmid R1/NR1. Further studies on the par regions of F and PI demonstrated a particularly high degree of similarity both in their genetic structure and the functionality of the two trans-acting proteins and a cis-acting site encoded by these two systems. Whereas the overall structure of the genetic region of the R1/NR1 systems showed some differences from the F and P1 systems, the analogies between the par regions of these IncFII plasmids and F/PL are striking. The identification of cis-acting stabilization regions in additional plasmids of Gram-negative bacteria soon followed, and the various par systems have been grouped according to the structure of the cis-acting (centromeric) region and homologies between the ParA(Pl)/SopA(F) proteins that exhibit ATPase activity and the ParB(Pl)/SopB(F) DNA-binding proteins. A review article by D. R. Williams and C. M. Thomas that appeared in 1992 provided an insightful comparison of par systems known at the time (162). Several years earlier S. Austin and K. Nordstrom collaborated on two reviews of plasmid partitioning that put a number of observations on plasmid partitioning in perspective (163, 164). Their treatment of plasmid partitioning included proposing models that could account for both the partitioning of plasmid DNA during cell division and the observed incompatibility between normally compatible plasmids that carry components of the same par region. An earlier review by R. Novick (39) drew attention to the possible role of both replication control and partitioning in incompatibility.

Not all par regions of plasmids of Gram-negative bacteria fit the F/PL paradigm of two trans-acting proteins acting at a single in cis site. A somewhat atypical par system is found for the IncP plasmids RK2 and R751. These plasmids encode KorB and IncC proteins that are homologous to the ParA and ParB proteins, but in this case the KorB protein binds to 12 sites and, thus, may be involved both in global regulation and partitioning (109). Considerable work by the laboratory of S. Cohen on the par region of plasmid pSC101, which does not contain genes that encode trans-acting proteins analogous to the classic par systems, importantly established the role of superhelicity in the stable maintenance of plasmids and provided evidence for a role of a replication complex, including host replication proteins in the partitioning event (165, 166). In this regard it is of interest that studies on bacteriophage lambda over a number years by G. Wegrzyn and K. Taylor have supported the notion of inheritance of a stable lambda DNA replication complex (167). It appears that a number of cellular factors will be found that play a role in the active process of plasmid DNA partitioning. The work of S. Austin has demonstrated the importance of DNA condensation by SMC proteins in stable plasmid maintenance (168).

Another important variation of the model F/P1/R1 par systems is the recent awareness of a basic class of plasmid replicons, designated RepABC, where both replication- and partitioning-like elements are present in a single operon (169). These plasmids, found mostly in Rhizobium and Agrobacterium characteristically consist of a replication initiation protein gene (repC) and two genes (repA and repB) that encode for proteins homologous to the family of ParA and ParB proteins. Because it is only recently that stable maintenance regions of the RepABC plasmids and plasmids of Gram-positive bacteria have been under study, it remains to be seen what similarities there are mechanistically between the activities of these somewhat atypical par regions and the F/P1/R1 system.

While much of the work on plasmid partitioning in the 1980s and 1990s characterized the biochemical properties of the various par components that provide for stable maintenance of low-copy-number plasmids, up until recently, conclusive evidence for a plasmid partitioning apparatus was elusive. This has now changed with the use of recently developed techniques for specifically tagging DNA molecules with green fluorescent protein by A. Belmont and A. Murray (170), which has led to the visualization of plasmid molecules in a growing population of bacteria (171) and importantly extended the study of plasmid DNA partitioning to include cell biology approaches. With this technique, plasmids F and PI have been localized to the midcell position in E. coli newborn cells, and immediately or sometime after plasmid duplication, they rapidly migrate to the one-quarter and three-quarter positions that mark the future midpoints of the nascent daughter cells (171). Similar results by the laboratory of S. Hiraga have been obtained by fluorescence in situ hybridization using fluorescent labeled probes to localize plasmid F (172). The findings of localization of plasmids F and PI and the dynamic movement of these plasmids suggest the presence of host structures that play a role in the localization of a plasmid and an active process of movement of the copies of plasmids after duplication to the future midcell position of nascent daughter cells. Localization and dynamic movement have also been demonstrated for plasmid Rl in E. coli by the K. Gerdes laboratory (173). Using immunofluorescence analysis, the laboratory of C. M. Thomas demonstrated the localization of the KorB protein, a homologue of ParB, to mid- and quarter-cell positions in E. coli carrying plasmid RK2 (174). Because KorB binds specifically to sites on RK2, this suggested a pattern of localization for this plasmid similar to that found for PI and F. Localization and actual movement of the multicopy plasmid RK2 by real-time fluorescence microscopy was demonstrated more recently with the green fluorescent protein-Lac tagging technique (175). Surprisingly in this same study, localization and movement of a pUC plasmid carrying lacO inserts were observed in a large portion of the population of E. coli cells, suggesting that plasmids can exhibit some degree of localization in the absence of a plasmid-encoded par system. Furthermore, the finding of one or a few foci per cell for multicopy RK2 and pUC plasmids clearly indicates a clustering of plasmid molecules at one or a limited number of sites.

Evidence has been obtained for the localization of the DNA replication machinery at the mid- and quarter-cell positions in both Bacillus subtilis and E. coli, suggesting stationary replication factories (176). These replication complexes at relatively fixed positions may be largely responsible for the localization of plasmid elements. It is additionally possible that the attachment of a plasmid element to the replication complex is stabilized by the plasmid par system, particularly during the process of movement of plasmids to new fixed positions (175). Colocalization of plasmids and partitioning proteins of F, PI, Rl, and RK2 is consistent with this notion. Using the fluorescence in situ hybridization probe technique, it was recently found that the following pairs of plasmids, F and Pi, RK2 and F, RK2 and PI, did not colocalize in E. coli but occupied separable positions in the mid- and quarter-cell locations (177). In the same study the localization of plasmid RK2 was extended to Pseudomonas aeruginosa and Pseudomonas putida with similar results as with E. coli. Conceivably, the observed movement of plasmid DNA occurs with associated Par proteins with or without host components to new sites at fixed positions. Studies on the mechanism of localization of clusters of plasmid molecules and the dynamic movement of these clusters represent an exciting new chapter of plasmid research. The finding of clustering of plasmids, rather than a random distribution of plasmid copies, also has major implications regarding the interplay of plasmid components involved in copy-number control. It is clear that the addition of cell biology approaches to the genetic and biochemical tools available for plasmid research has substantially enhanced prospects for understanding structures and events responsible for the complex processes of plasmid replication, partitioning, and conjugal transfer.

CONCLUDING REMARKS

The early investigators of the extrachromosomal and conjugal transfer properties of F, R, and Col plasmids had little or no access to the sophisticated set of tools that are now commonplace for the study of the structure of these elements. Despite limited resources, the findings that were made during this early chapter of plasmid research have largely held up under the scrutiny of later and more sophisticated molecular analyses. As an increasing number of plasmids were identified and characterized in both Gram-negative and Gram-positive bacteria, particularly with regard to the various genes and mobile elements that they possessed, it almost seemed at one point that plasmids represented an endless set of mosaic structures. Although superficially this was the case, as the number of reports on DNA sequences within these mosaics and the specific functions they encoded substantially increased, it became clear that the number of combinatorial structures of plasmids is likely not to be endless but is finite.

The similarities in mechanisms of regulation of replication, conjugal transfer, and translocation events across different bacterial species are striking. That is not to say that there are not significant differences between plasmids, particularly in comparisons of plasmids of Gram-positive and Gram-negative bacteria, but most plasmids can in fact be placed into a specific group on the basis of their overall properties, and the number of these groups or categories is not extremely large. This is also true for mobile elements. These similarities in structure and function in no way minimize the importance of the continued study of both previously characterized and new plasmid elements. We are still in the very early stages in our understanding mechanistically of a number of fundamental plasmid properties, including partitioning, conjugal transfer, the acquisition of whole sets of genes, and both environmental and bacterial host effects on plasmid gene expression and transfer during biofilm formation and during the course of invasion of mammalian hosts by pathogenic bacteria. The tasks ahead are somewhat daunting and scientifically challenging when one considers that we have only begun to tap the vast number of bacteria and their plasmids in our biosphere. Fortunately, new optical, structural, and nano technologies have been developed and they will continue to be improved. These developments, along with expected new technical breakthroughs, will open fresh opportunities for the analysis of complex events from a structural standpoint as, for example, conjugal transfer, plasmid localization and dynamic movement, and interbacterial communication and plasmid-encoded bacterial-eukaryotic host interactions. These same technological advances are also greatly enhancing our ability to examine plasmid movement in complex microbial communities in natural environments. Through the efforts of a number of investigators over the years, there is now a solid and growing base of plasmid properties that should with advancing technologies allow us to examine plasmid phenomena that we could only dream about tackling not so long ago. It is of course not possible to track the journey of a naturally occurring plasmid from the time it was born and as it acquired and lost genes and functions during its passage through time and through a number of different bacterial hosts that in turn were exposed to different environments. Nevertheless, the structural information that can now be obtained for a specific plasmid element certainly can give definite hints as to the journey that this plasmid has made and possibly the journey that is yet to come.