Linear plasmids and chromosomes in bacteria

Linear plasmids and chromosomes were unknown in prokaryotes until recently but have now been found in spirochaetes, Gram‐positive bacteria, and Gram‐negative bacteria. Two structural types of bacterial linear DNA have been characterized. Linear plasmids of the spirochaete Borrelia have a covalently closed hairpin loop at each end and linear plasmids of the Gram‐positive filamentous Streptomyces have a covalently attached protein at each end. Replicons with similar structures are more frequent in eukaryotic cells than in prokaryotes. Linear genomic structures are probably more common in bacteria than previously recognized, however, and some replicons may interconvert between circular and linear isomers. The molecular biology of these widely dispersed elements provides clues to explain the origin of linear DNA in bacteria, including evidence for genetic exchange between prokaryotes and eukaryotes.

This classical view must now be amended in view of a new round of examinations of the bacterial genome. Pulsed-field gel electrophoresis, used to separate large DNA molecules and to distinguish between circular and linear forms, has revealed that prokaryotes show a variety of genomic configurations. In 1989, the first linear bacterial chromosome was discovered in the arthropod-borne spirochaete Borrelia burgdorferi, the agent of Lyme disease (Baril etai, 1989;Ferdows and Barbour, 1989). Linear chromosomes have since been reported for Streptomyces coelicolor, Streptomyces lividans (Chen et ai, 1993a) and Rhodococcus fascians (Crespi et ai, 1992), which are members of the high G-HC subdivision of Grampositive bacteria. A multipartite genome consisting of one linear and one circular chromosome has been reported for Agrobacterium tumefaciens (A. Allardet-Servent, personal communication), a Gram-negative member of the a-group Protobacteriaceae. Other bacteria in this group have two distinct circular chromosomes, or one chromosome and one or more megaplasmids (Michaux et ai, 1993;Suwanto and Kaplan, 1989).
Bacterial linear plasmids were first described in Streptomyces rochei'm 1979(Hayakawa ef a/., 1979 and have now been detected in at least 10 other Streptomyces spp. as well as in the related bacteria R. fascians (Crespi etai, 1992) and Nocardia opaca (Kalkus et ai, 1990). Linear plasmids are found in all members of the genus Borrelia (Plasterk etai, 1985;Marconi etai, 1993). A linear plasmid has also been reported in the Gram-negative Thiobacillus versutus (Wlodarczyk and Nowicka, 1988) and the prophage of coliphage N15 is a linear plasmid (Svarchevsky and Rybchin, 1984).
The discovery of linear chromosomes and plasmids in bacteria raises several questions. What are their molecular structures and replication mechanisms and how do they compare with more familiar linear DNA models? Also, since circular DNA typifies most prokaryotes, how did linear DNA originate in certain bacteria and what is its biological significance?

Prokaryotic teiomeres
In characterizing a linear genome, it is reasonable to begin at its ends. In fact, proof of a linear genomic structure requires demonstrating unique molecular ends. A telomere is the region of DNA at the end of a linear replicon that is essential for the stability and complete replication of the molecule (Blackburn, 1991). One type of telomere, found on eukaryotic chromosomes from protozoa to humans, is composed of tandemly repeated, simple sequences (Blackburn, 1991). The prokaryotic linear repiicons that have been characterized exhibit two very different telomere types, reflecting different strategies for stabilizing and replicating the ends (Fig. 1).

Borrelia linear plasmids: hairpin teiomeres
Spirochaetes of the genus Borrelia have a predominantly linear genomic structure. All species that have been examined contain a 950-1000 kb linear chromosome and several linear plasmids that range in size from 5 kb to greater than 200kb (Baril etai, 1989;Ferdows and Barbour, 1989;Staihammar-Carlemalm et ai, 1990; M. S. Ferdows and A. G. Barbour, personal communication). Borrelias also contain circular plasmids, the prototypical form of bacterial plasmid. The physical map of the B. burgdorferi chromosome is linear (Casjens and Huang, 1993;Davidson etai, 1992), but the chromosomal teiomeres have not yet been isolated and analysed.
The structure of a 16 kb linear plasmid of B. burgdorferi has been characterized in the most detail (Hinnebusch and Barbour, 1991;Hinnebusch etai, 1990). At each end of this plasmid, the two DNA strands are connected and form a perfectly palindromic AT-rich terminal hairpin loop. A conserved 19 bp inverted repeat sequence is present at each end of the plasmid, comprising an inverted terminal repeat. All Borrelia linear plasmids analysed to date appear to have these telomeric features (Barbour and Garon, 1987;Kitten and Barbour, 1990).
The prophage of the unusual temperate coliphage N15 is similar in structure to Borrelia linear plasmids (Svarchevsky and Rybchin, 1984). N15, likeA., has a double-stranded linear DNA genome with cohesive ends. Its prophage, however, replicates extrachromosomally in E. coli as a 46 kb linear plasmid. The N15 prophage has palindromic terminal hairpin loops and a 28 bp inverted terminal repeat (Maiinin etai, 1992b).
All other examples of this telomeric structure are found in eukaryotic cells and their viruses. Borrelia linear plasmids most closely resemble the genomes of vaccinia and other animal poxviruses (Baroudy et al., 1982), African swine fever virus (ASFV) (Gonzalez et ai, 1986), and mitochondria of the yeast Pichia (Dinouel et ai, 1993). These genomes also have AT-rich single-stranded hairpin loops that covalentiy link the two DNA strands, and inverted terminal repeats. The mitochondrial DNA of Paramecium has an AT-rich hairpin loop at one end (Pritchard and Cummings, 1981).
Some other eukaryotic linear genomes have hairpin structures, but the nucleotide sequences of their terminal single-stranded loops have not been determined. This group includes viruses of the unicellular alga Chlorella (Rohozinski et ai, 1989), mitochondrial plasmids of the plant pathogenic fungus Rhizoctonia solani (Miyashita et ai, 1990), and deletion derivatives of the chloroplast genome of rice (Harada etai, 1992).

Streptomyces linear plasmids: invertron teiomeres
A different type of linear plasmid is found in several species of the filamentous soil bacteria Streptomyces. These plasmids, which range from 9 kb to more than 600 kb, have a terminal protein covalentiy attached to the 5' end of each DNA strand, as well as inverted terminal repeats (Sakaguchi, 1990), The Streptomyces chromosome may have a similar structure (Chen etai., 1993a;C, Chen and H. Kieser, personal communication). These telomeric features place the Streptomyces plasmids in a widely dispersed class of linear DNA replicons encountered mainly in eukaryotic cells (Fig. 1), Virtually every eukaryotic linear plasmid is of this type, including cytoplasmic, mitochondrial, and chioroplast plasmids of many fungi, algae, and higher plants (Meinhardt et ai., 1990), Adenoviruses, which infect animal cells, also have genomes with this type of telomere (Honwitz, 1990), Among prokaryotes, a few phages, including the Baciiius subtilis phage <|)29, Streptococcus pneumoniae phage HB-3, and E, coii phage PRDl have the same structure as the Streptomyces linear plasmids (Salas, 1988).
Several other features link these phylogenetically diverse elements. Although they have not yet been characterized for Streptomyces linear plasmids, the DNA poiymerases utilized by other members of this group are more closely related to eukaryotic DNA polymerase a than to prokaryotic DNA polymerase Poll (Salas, 1991). Many, including bacteriophage HB-3 (Romero et ai., 1990), several fungal and plant mitochondrial plasmids (Meinhardt etal., 1990), Adenovirus (Honwitz, 1990), and linear plasmid SCP1 of Streptomyces (Kinashi ef al., 1992) can integrate into the host chromosome. Sakaguchi (1990) proposed the term invertron for these autonomously replicating mobile genetic elements with covalently attached terminal proteins and inverted terminal repeats, and discussed their structural and functional similarities to transposons.

Replication of bacterial linear DNA
As first pointed out by Watson (1972), it is not obvious how the ends of a linear molecule can be completely replicated. Every DNA polymerase requires a pre-existing primer, usually a short RNA molecule, to provide a free 3'-OH group for initiating 5'-to-3'-directed growth. Once a linear molecule is primed at one end, the nascent strand can grow to the opposite end. However, primer removal leaves the daughter molecule with a 3' tail of unreplicated DNA, This dilemma is resolved by replication mechanisms that rely on special features of the telomeres.

Hairpin piasmid repiication
When the rules of enzymatic DNA polymerization became known, Cavalier-Smith (1974) and Bateman (1975) theorized that telomeres were palindromic hairpins and showed how these structures could provide a means of complete replication. Their models, originally proposed for eukaryotic chromosomes, precisely describe the telomeres of Borreiia plasmids. The Bateman Model predicts a concatemeric replicative intermediate, in which unit-length genomes are connected by duplex forms of the hairpin loop sequence. We have found evidence for such a concatemer junction, composed of the fused left and right ends of the Borreiia 16 kb linear plasmid (K. Tilly and J. Fuhrman, unpublished; J, Hinnebusch and A, G, Barbour, unpublished). Whether this junction fragment was derived from a replicative form of the Borreiia linear plasmid is not known, but similar concatemer junctions are present in replicative forms of vaccinia virus (Merchlinsky and Moss, 1986). A duplex form of the N15 prophage hairpin sequences also occurs in the middle of the encapsidated phage genome (Malinin etai., 1992a).
Although no replication origins have been identified for the Borrelia plasmids or chromosome, preliminary genetic mapping of its linear chromosome suggests two possible locations. The DNA replication genes dnaA, dnaN, gyrB, and gidA are near the chromosomal origins of other bacteria. In S. burgdorferi, the dnaA cluster is at the centre of the chromosome but the gidA gene is near the left end. Old etal. (1992) proposed that the replication origin is at one of these locations.

invertron repiication
Streptomyces linear plasmids and other invertrons probably replicate by the protein-primed DNA replication mechanism that has been well characterized for B. subtiiis phage $29 and Adenovirus (Salas, 1991), The telomere is the origin of replication and is recognized by specific initiation proteins that promote unwinding. The 5'-terminal protein (TP) serves as the primer for a specific DNA polymerase. In the priming reaction, a free TP first complexes with DNA polymerase at the telomeric origin. The DNA polymerase then catalyses the formation of a covalent bond between the free TP and a dNTP. This TP-linked dNTP becomes the 5' terminal nucleotide of the nascent strand, which can then grow to the end of its template.

Origin of linear DNA in bacteria
The Borrelia and Streptomyces plasmids have structures rarely described in prokaryotes, but replicons with these structures are widespread in eukaryotic cells and their organelles. There are three formal possibilities by which these similarities could have arisen: (i) descent from a common ancestor; (ii) convergence to fulfil similar functions, with no recent common ancestor; and (iii) 'borrowing', or a direct transfer from one organism to another (Diamond, 1990), At present there is no clear favourite among these, and more than one mechanism niay have operated to produce similar genetic entities in phylogenetically diverse prokaryotes and eukaryotes.
Linear plasmids could have evolved from bacteriophages, as postulated for circular plasmids, A common phage or eukaryotic virus ancestor has been proposed for all the invertrons (Meinhardt et al., 1990) and a few phages, such as (|)29, have the same structure as Streptomyces linear plasmids, Borrelias harbour bacteriophages, but it is not known if borreliaphages have hairpin genomes like the N15 prophage (Barbour and Hayes, 1986), Although many bacteriophage genomes are linear double-stranded DNA, they typically have free 5' and 3' ends or self-complementary, single-stranded terminal overhangs.
Genetic exchange between prokaryotes and eukaryotes may have played a role in the dispersal of linear replicons. Two candidates for such a frans-kingdom transfer are Borrelia duttoni and the animal virus agent of ASFV. Both of these arthropod-borne pathogens are transmitted by the tick Omithodoros moubata. The telomeric hairpin structures and sequences of the ASFV genome and Borrelia linear plasmids are similar, indicating that the Borrelia telomeres may have been acquired through a horizontal transfer from ASFV (Hinnebusch and Barbour, 1991).
The Streptomyces conjugative linear plasmids SLP2 and pBL1 can be transferred to different species (Chen et al., 1993b;Zotchev et al., 1992), and linear plasmids of related bacteria are transferred among different genera (Kalkus etal., 1990), They may have strayed even farther afield. The organization and nucleotide sequences of some Streptomyces antibiotic synthesis genes are similar to those of certain fungi (Hara et al., 1991). Genetic exchange between soil bacteria and fungi may account for this similarity (Weigel et al., 1988), and Kinashi et ai (1992) hypothesized that S/reptomyces conjugative linear plasmids were likely vehicles for this horizontal exchange.
Perhaps the most famous example of prokaryoteeukaryote gene exchange is the transfer of DNA from the phytopathogen A. tumefasciens into plant cells to induce crown gall tumour. A similar pathogenesis that involves a linear plasmid may occur with R. fascians, an unrelated plant pathogen. This Gram-positive organism causes leafy gall and contains a 4 Mb linear chromosome, as well as a 200 kb linear conjugative plasmid that carries the genes responsible for virulence (Crespi et ai., 1992).

Conversion between circular and linear forms
Since circular and linear molecules require different replication mechanisms, the two would appear to be evolutionarily distinct. However, a growing body of evidence suggests that linear variants can arise during replication of circular genomes. The genomes of mitochondria and chloroplasts seem to be particularly plastic in this regard. Circular and linear genomic isomers can coexist in the organelles of fungi, algae, and plant cells (Bendich and Smith, 1990), Mitochondria of the yeast Pichia have linear genomes with terminal hairpin loops, but circular monomeric forms are also present (Fukuhara et al., 1993). Conversely, the chloroplast genomes of barley and rice are ordinarily circular, but linear hairpin deletion derivatives have been isolated from albino clones (Ellis and Day, 1986;Harada et al., 1992). Fukuhara et al. (1993) found that species of two related yeast genera have either circular or linear mitochondrial genomes. Their physical and genetic maps are highly conserved, however, suggesting that the two forms diverged recently by a relatively trivial mechanism. There are simple models to explain the change from circular to linear DNA (Ellis and Day, 1986;Fukuhara etai, 1993), Conversion might also occur by the action of a mobile DNA element. A transposon composed of two joined telomeres and the gene for a telomere-resolving enzyme could convert a circular molecule to a linear form by integration, cleavage, and telomere formation.
Bacterial linear plasmids and chromosomes may undergo similar transitions. Recently, a stably inherited circular derivative of one of the Borrelia linear plasmids was detected (M, S, Ferdows and A. G. Barbour, personal communication), A Streptomyces linear plasmid has also been shown to be able to replicate in either a linear or a circular form (Shiffman and Cohen, 1992), and although the Streptomyces chromosome may be linear, the S. lividans chromosome has a typical oriC that functions on circular minichromosomes (Zakrzewska-Czenwinska and Schrempf, 1992). Consistent with these findings, homologues of bacterial circular DNA replication genes have been detected on a Streptomyces linear plasmid and the Borreiia linear chromosome (Old et aL, 1992;Wu and Roy, 1993). In addition, a change from one telomere type to another has been documented for linear replicons, Cytoplasmic linear plasmids of the yeast Kiuyveromyces are invertrons but, when introduced into Saccharomyces, they generate derivatives that have a hairpin loop at one end and a covalently bound terminal protein at the other end (Kikuchi etal., 1985).
Linearization of circular replicons and at least transient stabilization of the free DNA ends occurs during replication, mobilization of conjugative plasmids, and integration and excision of phages and transposons. Many bacteriophages, including X, have linear encapsidated genomes yet replicate as circles or concatemers (Furth and Wickner, 1983). Therefore, the distinction between circular and linear replicons may not always be absolute. Evidence for a linear genomic structure may often be dismissed as artefact but it may provide insight into the molecular biology of the replicon.

Significance of bacterial linear DNA
Whatever the origin of linear DNA in bacteria, this form may be maintained because of inherent advantages it provides. Teiomeres are often recombinogenic because of reiterated sequences. In lower eukaryotes telomeric rearrangements and translocations provide a means for the duplication and molecular evolution of telomere-associated genes and gene families (Charron ef ai, 1989). Recombination within a telomeric multigene family is responsible for most of the genetic differences seen among ASFV isolates (de la Vega et ai, 1990) and rearrangements among telomeric genes of poxviruses can extend host range (Buller and Palumbo, 1991). The Borrelia species that cause relapsing fever appear to have exploited this potential. These organisms escape the immune response of their mammalian host by switching expression among members of a family of antigenically distinct outer surface protein genes. These genes are located on linear plasmids and the promoter of the unique expression site is near a telomere. Antigenic variation results from the replacement of one gene with another at the telomeric expression site, a phenomenon very similar to outer-surface-protein gene switching in trypanosomes (Borst, 1991). Because they occur at the ends of linear molecules, these frequent rearrangements do not disrupt expression of any downstream genes (Kitten and Barbour, 1990).
Conjugation and chromosomal integration of Streptomyces linear plasmids and other invertrons may be simplified by their linear structure, since these processes generally require linear intermediates. In addition, possessing a type of conjugative element that is common to both prokaryotes and eukaryotes may allow genetic communication between actinomycetes, fungi, and plants in complex, competitive soil communities.

Future directions
Many questions about the molecular biology of bacterial linear plasmids and chromosomes remain unanswered. The teiomeres of linear bacterial chromosomes have not yet been analysed. Genes, proteins, and DNA target sequences required for replication and segregation are undefined. An important step will be developing in vitro DNA replication systems, and genetic and biochemical studies of Borrelia plasmids may be simplified if coliphage N15-derived chimeric linear plasmids function in E. coli.
Reports of linear plasmids and chromosomes in bacteria are increasing and evidence is beginning to grow that some repiicons may be maintained in both circular and linear forms. Understanding the molecular biology of linear genetic elements of bacteria and how these novel elements relate to circular repiicons will lead to a more complete picture of the bacterial genome.

Note added in proof
Further evidence for the linear structure of the Streptomyces lividans chromosome is presented in the paper by Lin etai (MolMicrobiol, 1993,10: 923-933-this issue).