Home | About Us | Publications| Links

Complications and implications of linear bacterial chromosomes

C. W. Chen

Trends in Genetics 12: 192-196 (1996)

The emergence of linear bacterial chromosomes has overthrown the dogma of universal circularity of the bacterial chromosomes, and posed mechanistic and evolutionary implications not previously anticipated. These are the subject of this article.

The first linear bacterial chromosome, that of the spirochete Borrelia burgdorferi, was readily detected by standard pulsed-field gel electrophoresis technique due to its relatively small size (960 kb)1. Then, there was evidence from physical mapping that one linear and one circular chromosome exist in Agrobacterium tumefaciens2. For the chromosomes of gram-positive bacteria Streptomyces, which are among the largest in bacteria (about 8 Mb), discovering and proving their linearity was significantly more difficult3. It was achieved first in Streptomyces lividans 66, and later all other species investigated3-6. Rhodococcus fascians, another member of the actinomycetes, also appeared to have a linear chromosome7. Application of the powerful new tools are likely to uncover linear chromosomes in more bacteria. The linear chromosomes of Streptomyces (particularly that of S. lividans), being best characterized, will be used to exemplify the discussion hereafter.

Linear bacterial chromosomes are often accompanied by linear plasmids. All the linear chromosomes and plasmids of Streptomyces characterized appear to contain terminal inverted repeat (TIR) and covalently-bound terminal proteins (TP). Although these features are shared by a group of linear replicons including some viral and plasmid DNAs, of which the most studied cases are adenoviruses and bacteriophage [phi]29 of Bacillus subtilis (reviewed in Ref. 8), there is at least one basic difference: while replication of adenoviruses and [phi]29 DNA is initiated at the end (using the TP as the primer) and proceeds through the whole genome to the other end, replication of linear plasmids9 and chromosomes10 of Streptomyces starts at an internal origin.

The sizes of the TIR range widely on the linear plasmids (0.04 to 80 kb)11, 12, and linear chromosomes (24 to 210 kb)3, 5, 6 in Streptomyces. At least one linear plasmid (SLP2 in S. lividans) shares a long (16 kb) terminal sequence with its chromosomal counterpart3, 11. The terminal sequence of the S. lividans chromosome is rich in palindromes (four elements each of 13-23 bp including a 3 bp gap, in the first 90 bp), a feature shared by the termini of the linear plasmids4 and chromosomes (C.-H. Huang and C. W. Chen, unpublished). A comparison of these terminal sequences reveals a strong tendency to preserve palindromicity: a base substitution in one repeat is usually accompanied by a complementary substitution in its partner sequence. Protein-binding palindromes frequently tolerate imperfection in the repeats, so the persistent palindromicity may reflect a different function (see later).

Linear plasmids of Borrelia, like those of Streptomyces, also range widely in size -- from 5 kb to > 200 kb. The termini of a 16 kb linear plasmid of B. burgdorferi contain a perfectly palindromic A+T-rich hairpin loop and a TIR13, a feature shared by other Borrelia linear plasmids. The termini of the Borrelia chromosomes have not been isolated, and their structures and sequences are unknown. Genetic organization resembling that of a typical bacterial oriC was also located in the middle of the B. burgdorferi chromosome.

Complications in replication telomere patching

Perhaps the most prominent issue is how a linear bacterial chromosome is replicated. The replication origin (oriC) of Streptomyces14, 15 and the putative oriC of Borrelia 16, 17 resemble those in other eubacteria in genetic arrangement, and there is no reason to doubt that they share the basic mechanism and regulation. The main issue is at the termini, which normal replicating machinery cannot fully replicate18.

Shiffman and Cohen19 showed that the linear plasmid pSCL of S. clavuligerus could replicate as a circular DNA molecule when the telomeres were removed, indicating the existence of an internal origin of replication within this linear replicon. Chang and Cohen9 further showed that replication of the linear plasmid pSLA2 of Streptomyces rochei proceeds from an internal origin toward the termini, leaving a 280-nt single-strand gap at the 3' end, which was supposedly patched up by TP-primed replication. Musialowski et al.10 showed evidence for replication of S. coelicolor initiated at oriC ran bi-directionally toward the telomeres. This would also leave a single-stranded gap at each end (Fig. 1a). How are the gaps filled?

FIGURE 1. Models for patching the telomeres of Streptomyces plasmids and chromosomes. (a) The putative telomere structure after replication. The upper daughter DNA is fully duplicated, whereas the lower one remains single-stranded at the 3' end. The four terminal palindromes (I - IV) in S. lividans are indicated. (b) The unlikely mechanism of patching replication by TP and DNA polymerase initiated at the single-strand 3' end. (c) Alternative model 1: TP and DNA polymerase recognize and initiate replication at the duplex formed by foldback of the protruding 3' end. (d) Alternative model 2: terminal initiation occurs at the fully duplex daughter DNA, displacing the TP-capped 5' strand, which is transferred by homologous recombination to the terminal 3' gap on the sibling chromosome. (e) Alternative model 3: the terminal palindrome of the protruding 3' end folds back and provides a primer for the patching synthesis, followed by nicking by TP and fill-in synthesis that restores the telomere. See text for details. A straight-forward solution is for DNA polymerase to start replication at the telomere using TP as the primer (Fig. 1b). There is, however, an important difference between this scheme and the replication of adenoviruses and [phi]29: The signal for the initiation of terminal replication in adenoviruses and [phi]29 is double-strand DNA, which provides sequence-specific structural information for recognition by TP and DNA polymerase8. Such a stable conformation is unlikely to be achieved by the single-strand terminus on the Streptomyces chromosomes in the form of a random coil. Other features of the terminal DNA may be invoked.

There are at least three alternatives. In the first model (Fig. 1c; initially suggested by S. N. Cohen), the protruding 3' end folds back and pairs with an internal sequence. The abundance and persistence of palindromic sequences in the terminal regions support this view. For example, the terminal 13 bases (palindrome I) of the S. lividans chromosome may fold back and pair with a complementary sequence in palindrome IV (59 bp away)4. Formation of the I-IV duplex may be aided by fold-back of the intervening palindromes II and III. A patching replication is initiated at the I-IV duplex by the TP and polymerase and proceeds inward along the single-strand 3' end.

In the second model (Fig. 1d), the TP primes replication at the telomere of the fully duplicated daughter DNA, displacing the parental 5' strand (with the attached TP), which then pairs with the protruding 3' end. Strand exchange and resolution of the half-Holiday junction is then carried out by homologous recombination. This model, therefore, suggests an essential role of recA (for strand invasion) and ruv genes (for resolution) in cell viability. Indeed, there has been unexpected difficulty in generating null mutations of recA in Streptomyces using conventional gene targeting procedures (A.-J Cheng, S.-T. Hu, C. W. Chen, unpublished results; W. Wohlleben, personal communication). Moreover, strand exchange in this model may be across two arms of the replicon, which would allow homogenotization of the terminal sequences or repair of a damaged telomere through gene conversion.

In the third model (Fig. 1e), palindrome I folds back on itself to form a hairpin such that its 3' end serves as the primer for gap filling replication. The TP nicks the resulting hairpin at a site opposite the original 3' end, and becomes covalently bound to the 5' end of the nicked strand. Further extension synthesis from the new 3' end at the nick unfolds the hairpin and restores the telomere structure. This mode of terminal patching is analogous to the 'rolling-hairpin' model for replication of parvovirus DNA18, in which a viral-encoded protein carries out site-specific incision while remaining covalently linked to the 5' end. This type of endonucleolytic protein also includes the gA protein of [phi]X174 phage, the g2 protein of Ff phage and the TraI protein of F plasmid. In this model the chromosomal TP has a different role than the replication-priming TPs of adenoviruses and [phi]29. Moreover, this model predicts that each round of replication will flip the terminal palindrome sequence once. This leads to a testable prediction that, among the population of S. lividans chromosomes, the imperfect Palindrome I exists in both orientations. This kind of 'hairpin transfer' has been demonstrated in parvovirus replication. The linear chromosomes of Borrelia, if they contain hairpin termini like their linear plasmid counterparts, may use the same mechanism proposed for other hairpin-capped DNA such as poxvirus and ridovirus20, which also shares the 'hairpin transfer' mechanism although the endonuclease does not remain covalently attached.

Complications in replicative transposition unresolvable cointegrates

Replicative transposition between linear bacterial replicons also has intriguing complications. In the traditional model built on circular replicons, replicative transposition generates a circular cointegrate molecule containing donor and target molecules separated by two directly repeated copies of the transposable element (Fig. 2a). For one class (such as Mu phage of E. coli), there is no resolution system, and the cointegrate is the final product. For another class of transosable elements typified by the Tn3 family, the cointegrate is resolved by an element-specific resolvase that aligns the resolution (res) sequences on the elements and carries out site-specific recombination21. There is a barrier to intermolecular resolution (leading to fusion of the circular replicons); no such reaction has been reported.

Application of these models of replicative transposition to linear replicons present a problem. While the recipient DNA would receive a new copy of the transposable element, two separate molecules (instead of a fused cointegrate) would be produced, which are unlikely to be 'resolved' even with a resolution system (Fig. 2b). Thus, the two recombinat replicons would conceivably remain the final product. In such transposition between a linear chromosome and a linear plasmid, a segment of the chromosome would be transferred to the plasmid. This may be a serious hazard if the lost segment contains essential genes. If the transposition is between two chromosomal DNA molecules, the process is similar to an unequal crossover, producing a deletion in one chromosome and an insertion in the other. The scarcity of important genes in the terminal regions of the Streptomyces chromosomes3 implies that if replicative transposition is not to produce deletions that have serious effects, it could be targeted only to the terminal regions of the chromosome. In this regard, the existence of the partial SLP2 DNA segment at the end of the S. lividans chromosome bordered by a transposon11 (that is not present elsewhere in the chromosome) resembles such a transposition. The recent observation that a linear plasmid appears to integrate and replace the end of the Streptomyces rimosus chromosome22 offers an new opportunity to test the model.

FIGURE 2. Replicative transposition and replicon topology. (a) Circular replicons. Replicative transposition gives rise to a cointegrate, in which the two replicons are fused by the duplicated copies of transposable element. The element-specific resolvase aligns the res sites and resolves the two replicons by site-specific recombination. The continuity of the cointegrate molecule is important for the resolution reaction. (b) Linear replicons. Replicative transposition creates two recombinant replicons, each with a copy of the transposable elements at the junction of the exchange. This discontinuous 'cointegrate' may be the final product, because resolution is unlikely (shown in parentheses) in the absence of continuity between the two molecules, according to the current model.

Complications in conjugal transfer the terminal enigma

Conjugal transfer of plasmids or chromosomes has also been modeled on circular replicons18. In the classical model of the F plasmid of E. coli, transfer is initiated by nicking at a specific site (oriT) on the plasmid, followed by rolling-circle replication (Fig. 3a). The displaced 5' strand, covalently linked to a plasmid-specific transfer protein (TraI), is transferred to the recipient cell. There is no basic biochemical difference between transfer of the plasmid (F) and the plasmid-chromosome cointegrate (Hfr), except for their sizes.

A biochemically analogous model for the transfer of linear Streptomyces plasmids and chromosomes is for TP-primed replication to initiate at the telomere (oriT) and displace the parental 5' TP-capped strand into the recipient (Fig. 3b). This proposed conjugal replication, unlike the proposed terminal patching (Fig. 1), is initiated at the fully duplexed terminus. The conjugal replication proceeds along the entire chromosome, preempting the normal bi-directional replication as shown in E. coli. Moreover, a linear plasmid may provide trans-acting functions for the mobilization of a chromosome without physical integration. This agrees with the recent observation of Pettis and Cohen23 that chromosomal gene transfer in S. lividans required trans-acting but not cis-acting plasmid functions.

FIGURE 3. Conjugal transfer of circular and linear replicons. (a) Circular plasmid (F) and circular chromosome (Hfr) in E. coli: in both cases, F-specific proteins carry out a nicking-priming event at oriT on F DNA, where the 5' end is covalently attached to one of the transfer proteins (TraI), and a rolling circle replication is initiated at the 3' end. The protein-capped 5' strand is continuously displaced and transferred to the recipient cell. In Hfr cells, the chromosomal DNA is led by the F sequence into the recipient cell as an exogenote. (b) Linear plasmids and linear chromosomes of Streptomyces. The TP-capped ends may act as the origin of transfer (oriT), where the putative TP-primed replication is initiated, and the 5' TP-bound strand is displaced and transferred, as for F/Hfr transfer.


This article raises some of the issues inherent in the linearity of bacterial replicons, and offers models that suggest future investigations. It echoes the sentiment that what is true for E. coli may not necessarily be true for other bacteria. At present we know neither how many forms of linear chromosomes there are, nor how the eubacterial kingdom is divided with respect to chromosomal topology. Each type of chromosome may occupy a distinct taxonomic domain, which would imply that chromosome topology plays an important role in the evolution of bacteria. On the other hand, it is also possible that topological conversion is frequent, and that both linear and circular chromosomes exist in closely related bacteria. The long observed structural instability (extensive deletions and amplifications) of Streptomyces chromosomes has been recently attributed to rearrangements at the terminal regions, some of which even lead to circularization of the chromosome3. The significance of the structural instability of Streptomyces chromosomes has recently been discussed elsewhere24. It remains to be seen whether the topological conversion observed in Streptomyces3, 19 and mitochondria25 has an evolutionary role.


This work is supported by research grants (NSC84-0418-B010-033-BC, NSC85-2331-B010-005-BC) from National Science Council, Taiwan. I thank Drs. David A. Hopwood and Stanley N. Cohen for critical reading and helpful comments on the manuscript. This paper is dedicated to the memory of my father, C.-C. Chen.


1 Ferdows, M.S. and Barbour, A.G. (1989) Proc Natl Acad Sci U.S.A 86, 5969-5973
2 Allardet-Servent, A. et al. (1993) J. Bacteriol. 175, 7869-7874
3 Lin, Y.-S., Kieser, H.M., Hopwood, D.A. and Chen, C.W. (1993) Mol. Microbiol. 10, 923-933
4 Chen, C.W. et al. (1994) Actinomycetologica 8, 103-112
5 Lezhava, A.L. et al. (1995) J. Bacteriol. 177, 6492-6498
6 Leblond, P. et al. (1996) Mol. Microbiol. In press.
7 Crespi, M. et al. (1992) EMBO J. 11, 795-804
8 Salas, M. (1991) Annu. Rev. Biochem. 60, 37-91
9 Chang, P.C. and Cohen, S.N. (1994) Science 265, 952-954
10 Musialowski, M.S. et al. (1994) J. Bacteriol. 176, 5123-5125
11 Chen, C.W. et al. (1993) Mol Microbiol 7, 925- 932
12 Kinashi, H., Shimaji-Murayama, M. and Hanafusa, T. (1991) Plasmid 26, 123-130
13 Hinnebusch, J. and Barbour, G. (1991) J. Bacteriol. 173, 7233-7239
14 Zakrzewska-Czerwinska, J. and Schrempf, H. (1992) J Bacteriol 147, 2688-2693
15 Calcutt, M.J. and Schmidt, F.J. (1992) J. Bacteriol. 174, 3220-3226
16 Casjens, S. and Huang, W.M. (1993) Mol Microbiol 8, 967-980.
17 Old, I.G., MacDougall, J., Girons, I.S. and Davidson, B.E. (1992) FEMS Microbiol Lett 99, 245-250
18 Kornberg, A. and Baker, T.A. (1992) DNA replication, W. H. Freeman
19 Shiffman, D. and Cohen, S.N. (1992) Proc. Natl. Acad. Sci. USA 89, 6129-6133
20 Merchlinsky, M. (1990) J. Virol. 64, 3437-3446
21 Stark, W.M., Roocock, M.R. and Sherratt, D.J. (1989) Trends Genet. 5, 304-309
22 Gravius, B. et al. (1994) Microbiol 140, 2271-2277
23 Pettis, G.S. and Cohen, S.N. (1994) Mol. Microbiol. 13, 955-964
24 Chen, C.W. (1995) Trends Biotechnol. 13, 157-160
25 Schardl, C.L., Lonsdale, D.M., Pring, D.R. and Rose, K.R. (1984) Nature 310, 292-296