The Streptomyces wild-type chromosome is linear in all examples studied. The ends of the chromosome or telomeres consist of terminal inverted repeats of various sizes with proteins covalently bound to their 5′ ends. The chromosome is very unstable and undergoes very large deletions spontaneously at rates higher than 0.1% of spores. Frequently, the telomeres are included in the deletions. Loss of both telomeres leads to circularization of the chromosome. The wild-type chromosome can also be circularized artificially by targeted recombination. Spontaneously or artificially circularized chromosomes are even more unstable than the linear ones. High-copy-number tandem amplifications of specific chromosomal regions are frequently associated with the deletions. RecA seems to be involved in the amplification mechanism and control of genetic instability.
Spontaneous and induced genetic instability in Streptomyces
The genus Streptomyces is an important group of Gram-positive filamentous bacteria from soil, which produce a great variety of secondary metabolites, including about 60% of known antibiotics. They exhibit a complex (for bacteria) cycle of morphological differentiation: a substrate mycelium grows from the spore and differentiates into aerial mycelium, which, after septation, gives rise to chains of spores. Most of the wild-type (WT) strains cultivated under laboratory conditions produce pigments, some of which are antibiotics. An examination of the culture plates usually reveals some variants showing a modification or an absence of pigmentation, aerial mycelium or spore formation. The mutants arise at very high frequencies (more than 0.1% of colony-forming spores). This phenomenon of genetic instability seems to be ubiquitous in the genus Streptomyces (Leblond and Decaris, 1994; Dharmalingam and Cullum, 1996). Genetic instability affects different phenotypical properties, often pleiotropically, including morphological differentiation, production of secondary metabolites, such as pigments and antibiotics, antibiotic resistance, secretion of extracellular enzymes and sometimes genes for primary metabolism, particularly one or more steps in the arginine biosynthetic pathway. Usually, mutants generated by genetic instability are ‘hypervariable’, giving rise to new mutants without any preponderant phenotype (Leblond and Decaris, 1994). In some species, a sequential occurrence of mutations has been observed. In Streptomyces lividans, for example, chloramphenicol-sensitive (CmlS) mutants arise with a frequency of 0.5%, and they give rise to about 25% of arginine-auxotrophic (Arg−) mutants (Altenbuchner and Cullum, 1984). Mutations can occur at different stages of the colonial development. For example, in Streptomyces ambofaciens, mutations affecting colony pigmentation can arise during growth of the substrate mycelium or in the aerial mycelium, forming so-called sectors and papillae respectively (Leblond and Decaris, 1994; Vandewiele et al., 1996).
Genetic instability can be stimulated by mutagens, such as mitomycin C, ultraviolet (UV) light, nitrous acid and ethidium bromide (e.g. Volff et al., 1993a). The effect of antibiotics that inhibit bacterial topoisomerase II gyrase is particularly spectacular: at very high survival rates, the frequency of mutants can reach values close to 100% (Volff et al., 1993b).
Structure of the chromosome
The mycelium of Streptomyces is coenocytic, each hyphal compartment containing numerous copies of the chromosome. In contrast, at least the majority of the spores contain a single chromosome. The DNA is G+C-rich (72% on average). Streptomyces are one of the few examples of bacteria with a linear chromosome (Chen, 1996; Kolstø, 1997). The chromosomes of the Streptomyces species so far investigated are linear molecules of about 8 Mb with proteins covalently bound at the 5′ ends (Lin et al., 1993; Lezhava et al., 1995; Leblond et al., 1996; Pandza et al., 1997). Replication proceeds bidirectionally from an origin of replication oriC located in the centre of the chromosome (Fig. 1; Zakrzewska-Czerwinska and Schrempf, 1992; Musialowski et al., 1994). This would leave a single-stranded gap at each chromosome end, which can be filled by several mechanisms involving a DNA polymerase and terminal protein (Chen, 1996). The DNA sequence of one end of the chromosome (and of some linear plasmids) is inversely repeated at the other end. These terminal inverted repeats (TIRs) have sizes between 24 kb (Streptomyces griseus; Lezhava et al., 1995) and 550 kb (Streptomyces rimosus; Pandza et al., 1997). Many short palindromic repeats are present within the last few hundred basepairs of the TIRs (Lin et al., 1993).
About 17 kb of the extreme ends of the S. lividans chromosome are similar or identical to one end of the linear plasmid SLP2 (Lin et al., 1993). This includes one copy of the transposable element, Tn4811, which is present near the middle of the S. lividans 30 kb chromosomal TIRs. Another transposable element, IS1372, was recently identified in the TIRs 334 bp away from Tn4811 (Fischer et al., 1996). Two open reading frames encoding putative products belonging to the two-component signalling family lie close to the TIRs, one on each side. In addition, numerous long direct repeats are present within a region of about 400 bp next to one TIR (Volff et al., 1997a). No housekeeping genes, i.e. genes encoding functions involved in transcription, translation, DNA replication and central catabolic and anabolic metabolism, were located within about 800 kb from one and at least 300 kb from the other chromosomal end in S. lividans. In S. coelicolor A3(2), no gene essential for growth on minimal media except argG was mapped within 1.2 Mb of either end (Redenbach et al., 1996). Similar large regions, which appeared to be genetically ‘silent’, were also found at the ends of other Streptomyces chromosomes.
Most mutants arising by genetic instability have DNA rearrangements within the ends. Usually, the chromosomal ends are joined to form a circular chromosome, and the telomeres and parts of the flanking chromosomal regions are deleted (Lin et al., 1993; Redenbach et al., 1993; Lezhava et al., 1997). The WT chromosome can also be circularized artificially by targeted recombination (Lin et al., 1993; Volff et al., 1997b). Hence, the chromosome of Streptomyces can exist and replicate in either a linear or a circular form.
Large chromosomal deletions
Deletions are nearly always responsible for the mutant phenotypes arising by genetic instability. In S. lividans, CmlS and Arg− phenotypes result from deletion of a chloramphenicol resistance marker, cmlR, and of the arginosuccinate synthetase gene, argG, respectively (Betzler et al., 1987; Dittrich et al., 1991). In S. glaucescens, the hydroxystreptomycin phosphotransferase and the tyrosinase structural genes are included in the deletions (Birch et al., 1990). In S. griseus, the afsA gene, which encodes a protein essential for A-factor production, is deleted in some A-factor-negative mutants (Lezhava et al., 1997). As determined by pulsed field gel electrophoresis and chromosome walking experiments, deletions can be very large (up to 2 Mb in S. ambofaciens; Leblond and Decaris, 1994). Typically, more than 10% of the Streptomyces chromosome is dispensable under laboratory conditions. In most cases, one or both telomeres are included in the deletion (Lin et al., 1993; Redenbach et al., 1993; Fischer et al., 1997; Lezhava et al., 1997). In S. lividans, markers such as cmlR, located close to one telomere, are the first to be lost during the sequential deletion process that later remove other genes, such as argG. Nevertheless, large internal deletions that did not remove any chromosomal end have been observed in S. ambofaciens (Fischer et al., 1997). Important changes in protein expression were detected in rearranged mutants of the same species using two-dimensional electrophoresis (Dary et al., 1993). Hence, large deletions probably include numerous but non-essential genes and cause the observed pleiotropy of the mutants arising by genetic instability. On the other hand, some mutants show no detectable rearrangements. For example, about 99% of S. rimosus mutants affected in sporulation, pigmentation and oxytetracycline production showed no large-scale DNA rearrangements (Gravius et al., 1993); no large rearrangements were found among pigment-negative mutants derived from papillae in S. ambofaciens (Vandewiele et al., 1996); and in S. lividans, some CmlS mutants had no visible rearrangements, either in the unstable region or in the cmlR gene (Volff and Altenbuchner, 1997a; Volff et al., 1997b). Thus, genetic instability is not strictly correlated with large deletions.
Possible mechanisms of deletion
The discovery of the linearity of the chromosome and of the high instability of its telomeres (Lin et al., 1993; Redenbach et al., 1993) was an important step towards understanding genetic instability. There are several explanations for the high instability of the telomeres. For example, an intramolecular transposition of a transposon, such as Tn4811, with a replicative mode of transposition from one TIR to a region somewhere on the other end of the chromosome would lead, in one orientation of transposon insertion, to deletion of both chromosomal ends and to circularization of the chromosome (Fig. 1). Intermolecular transposition between chromosomes would cause an exchange of the TIRs (Chen, 1996) and generate internal deletions and DNA duplications (Fig. 1). Indeed, the presence of the terminal half of the S. lividans chromosomal TIR, including Tn4811, on the linear plasmid SLP2 probably reflects such an intermolecular transposition event. On the other hand, sequence analysis of junctions after large deletions in pleiotropic mutants of S. glaucescens and S. lividans gave no hint of the involvement of a transposable element but instead indicated illegitimate recombination events at microhomologies in some cases (Birch et al., 1991; Piendl et al., 1991).
Another explanation could involve a lack of sufficient terminal proteins in fast-growing strains, allowing access by exonucleases at the unprotected chromosomal ends. Such chromosomes could be rescued by circularization. So far, no such gene encoding a terminal protein has been isolated to test this hypothesis.
A further model (Fig. 1; Volff and Altenbuchner, 1997a; Volff et al., 1997b) is derived from that proposed for the repair of collapsed replication forks in Escherichia coli (Kuzminov, 1995). A replication fork moving from the centre of the molecule will collapse when it reaches a single-strand break. This collapsed fork may be repaired by homologous recombination allowing replication to continue. However, intramolecular illegitimate recombination events between a free end and a sequence at the other end of the chromosome would delete both telomeres and circularize the chromosome. Similarly, recombination of the free DNA end from the collapsed fork with a sequence upstream or downstream from the collapse point would give rise to a chromosome with telomeres but containing a duplication or an internal deletion (not shown in Fig. 1), as described in S. ambofaciens (Fischer et al., 1997). Accordingly, a reduction in the level of homologous recombination should lead to a decrease in repair and to an increase in the deletion frequency. In fact, a S. lividans mutant expressing a truncated RecA protein and having a strongly reduced but not totally abolished homologous recombination activity (Muth et al., 1997) segregated 70 times more chloramphenicol-sensitive mutants than the WT strain because of an enhanced frequency of deletions (Volff and Altenbuchner, 1997a). Complete absence of RecA might lead to fragmentation of the chromosome, so that a basal level of homologous recombination may be essential for viability. This may explain why extensive efforts by several groups to delete recA from the chromosome have been unsuccessful (Muth et al., 1997). Furthermore, as proposed for E. coli (Kuzminov, 1995), excision repair of lesions caused by DNA-damaging agents generates single-stranded regions and increases the collapse of replication forks. This may explain why mutagenic agents, such as UV and mitomycin C, stimulate genetic instability in Streptomyces. Genotoxic agents, such as gyrase inhibitors, may also cause the replication forks to stall and become destabilized, thus increasing collapses followed by deletions.
High-copy-number DNA amplifications
Large deletions arising by genetic instability are frequently associated with DNA amplifications. Amplified fragments are reiterated in tandem up to several hundred copies per chromosome and can represent up to 50% of total genomic DNA. Amplifications usually occur, without apparent selection, from specific chromosomal regions called amplifiable units of DNA (AUDs). AUDs are mainly located in unstable chromosomal regions and can also be partially or completely deleted. In mutants with amplifications, the deletions usually end near or within the amplified DNA.
Type II AUDs have two direct repeats of at least 0.8 kb. Typically, independent mutants carry the same amplified fragment, including one repeat and the sequence between the two repeats. The first type II AUD was found in a tylosin non-producing mutant of S. fradiae. This AUD is bounded by 2.2 kb direct repeats within the tylosin biosynthesis gene cluster (reviewed by Baltz and Seno, 1988). The non-producing phenotype was caused by a deletion accompanying the amplification, which removed tylosin biosynthesis genes. Type II amplifications were later found in other Streptomyces species too. In Streptomyces achromogenes ssp. rubradiris, amplification of AUD-Sar1 carrying a spectinomycin resistance gene resulted from selection for higher spectinomycin resistance (Hornemann et al., 1993). The 90 kb AUD2 sequence is another type II amplifiable unit found in S. lividans. AUD2 includes mercury resistance genes and is delimited by two identical direct repeats (Eichenseer and Altenbuchner, 1994). These repeats correspond to the functional insertion sequence, IS1373, suggesting that AUD2 may be a giant mercury resistance transposon (Volff and Altenbuchner, 1997b). However, the best-studied example of type II amplification is probably the AUD1 element of S. lividans, which is found as a highly reiterated 5.7 kb fragment in 90% of the CmlS Arg− mutants arising by genetic instability (Altenbuchner and Cullum, 1984). AUD1 has a compound structure composed of three 1 kb and two 4.7 kb repeats alternating in the same orientation (i.e. 1 kb − 4.7 kb − 1 kb − 4.7 kb − 1 kb). The two 4.7 kb repeats are identical and encode a putative product with a low similarity to chitinases (Volff et al., 1996). The 1 kb repeats encode DNA-binding proteins of the LacI/GalR repressor family, which regulate a flanking α-glucosidase operon (Piendl et al., 1994; Altenbuchner et al., 1995).
Requirements for DNA amplification
S. lividans AUD1 is the only element so far analysed for its requirements for efficient high-copy-number amplification. Various combinations of the sequences composing AUD1 were inserted into the circular, low-copy-number SCP2 plasmid and analysed in mutants with a spontaneous deletion of the whole chromosomal AUD1 sequence (Volff et al., 1996). Both 4.7 kb repeats could be mutated by introducing frameshifts, reduced in size to a common 2.7 kb segment or completely replaced by repeats of equal size constructed with E. coli DNA, without affecting the ability to amplify. However, when these repeats were further reduced in size, no amplification was detected. This indicated that a minimal length of direct repeats is necessary for AUD1 amplification to occur and suggested that homologous recombination is involved. Indeed, AUD1-derived amplifiable units did not amplify in the S. lividans recA mutant described above (Volff and Altenbuchner, 1997a). The ability to amplify was restored by reintroduction of the wild-type recA gene, suggesting that the recombination event between the two direct repeats, leading to type II amplification, is mediated by RecA.
The righthand 1 kb repeat (RDR) is also needed for amplification of AUD1 (Volff et al., 1996). ORF-RDR (encoded by RDR) binds to two short homologous but non-identical palindromic sequences, one upstream and one downstream of RDR. In experiments using SCP2, efficient amplification was achieved only when both binding sites and a functional RDR were present. Constructions without the downstream binding site showed strongly reduced amplification efficiency with RDR. The left and middle direct repeats (LDR and MDR respectively) differ from RDR at 63 nucleotide positions. Gel mobility shift assays showed that ORF-MDR binds to the same palindromic DNA sequences as ORF-RDR. CmlS Arg− mutants isolated from a strain in which RDR was disrupted, but retaining LDR and MDR and all binding sites, showed a normal amplification of AUD1. This suggested that LDR and/or MDR promote amplification as well.
Other AUDs, called type I AUDs, belong to the same chromosomal region but are heterogeneous in their size and end-points and can sometimes overlap each other. Amplifications deriving from these AUDs therefore differ in independent mutants. Type I AUD regions have been found in S. glaucescens (Birch et al., 1990), S. ambofaciens (Leblond and Decaris, 1994; Aigle et al., 1996) and S. lividans (Rauland et al., 1995; Betzler et al., 1997). The mechanism of type I amplification remains unclear. In S. ambofaciens, a mutant with a type I amplification overproduced RecA (Aigle et al., 1997), so RecA could play a role in type I amplification as well. On the other hand, this overproduction could be a consequence rather than a cause of amplification, which, at least in S. lividans, leads to production of single-stranded DNA (unpublished results). Furthermore, AUD6 of S. ambofaciens, classified as a type I element, contains two nearly identical copies of the 1 kb repeat (Aubert et al., 1993). This might suggest that DNA-binding proteins involved in AUD1 type II amplification may also play a role in some type I amplifications.
Finally, the location of the amplifiable region might play a role in determining if and how a DNA fragment will be amplified. In a spontaneous CmlS Arg− mutant with a complete deletion of the chromosomal AUD1, the amplifiable unit was reintroduced into the chromosome at the integration site of the temperate phage φC31, which lies near the replication origin in the closely related strain S. coelicolor A3(2) (Redenbach et al., 1996). Here, no or a very low amplification of AUD1 was seen, indicating that the position of the AUD is important for its amplification (unpublished results).
Circular vs. linear chromosomes
To test a possible link between chromosome linearity and instability, the S. lividans chromosome was circularized by homologous recombination in an event that deleted part or all of the TIRs. Mutants with a circular chromosome retaining more than 10 kb of each 30 kb-long TIR showed phenotypical disadvantages, such as slower growth and reduced sporulation (Lin et al., 1993). In contrast, strains in which the chromosome was circularized with concomitant deletion of the TIRs showed no visible phenotypical disadvantage compared with the WT strain. However, in the latter case, these strains segregated about 20 times more CmlS mutants than the WT (Volff et al., 1997b). This was because of more frequent large deletions and is in good agreement with the fact that spontaneous CmlS mutants of S. lividans, which all or nearly all have a circularized chromosome, are also very unstable and segregate Arg− mutants at a frequency of about 25%. Similarly, pigment-negative mutants of S. ambofaciens with a circular chromosome turned out to be very unstable as well (Fischer et al., 1997). The region affected by deletions corresponds to the place at which the two replication forks should meet. Possible explanations for the very high instability of Streptomyces circular chromosomes are inefficient decatenation, lack of replication terminator sequences or problems in anchoring and separating the newly replicated chromosomes.
When the S. lividans TK64 chromosome was circularized by targeted recombination, amplifications of different sizes were observed at the fusion point in all strains tested. Regions of about 40–60 kb were amplified to about 10 copies per chromosome; they were rapidly lost by deletion and replaced by the classical AUD1 amplification in CmlS Arg− mutants. In S. lividans ZX7, an unstable amplification of a 20 kb sequence was found at the new chromosome junction (Chen et al., 1994). Here again, the amplified sequence could correspond to the region where the two replication forks meet. S. lividans AUD1 is mainly amplified in CmlS Arg− strains whose CmlS progenitors have a circular chromosome. Finally, the vector used for the extrachromosomal amplification system in S. lividans is circular (Volff et al., 1996). Hence, there seems to be no doubt that DNA amplification can arise in circular molecules. In vitro and in vivo replication experiments in other bacteria have shown that the lack of terminators leads to DNA over-replication and instability (Hiasa and Marians, 1994; Krabbe et al., 1997), so amplifications in Streptomyces circularized chromosomes may be caused by the absence of replication terminators.
Because AUD1 is not amplified in the S. lividans WT chromosome, the question arises whether amplification can occur in a linear molecule. Apparently, it can, because some S. lividans and S. ambofaciens mutants carrying amplifications seem to have a single TIR deleted, suggesting a linear structure (Rauland et al., 1995; Fischer et al., 1997; Volff et al., 1997b). It is unlikely that strains would survive with a chromosomal end not protected by the terminal protein. Maybe the undetected TIR has suffered an internal deletion, leaving just the extreme end of the TIR, which could still allow binding of the terminal protein but might be too short to be detected in Southern blots. Alternatively, a secondary site may be used for binding of the terminal protein after deletion of the normal binding site. As most of these strains had DNA amplifications, a model was proposed in which a rolling circle was formed by recombination on the side of the chromosome that had undergone deletion (Rauland et al., 1995). This would protect the end and generate amplification by over-replication. A problem with this model is how such chromosomes would separate after a new round of replication. We favour another explanation (Volff et al., 1997b), in which a chromosome with a deletion at one end could fuse via this end, in inverted orientation, with a second chromosome and thereby delete the same chromosomal end in the second chromosome (Fig. 1). This new chromosome, consisting of two inverted fused chromosomes, would have the same chromosomal end on each side, with a bound terminal protein. In Southern blots, it would appear as a single chromosome with one end deleted. In some of these chromosomes with two origins of replication, two replication forks could meet in the middle of the molecule. As postulated for circular chromosomes, this could induce rearrangements and generate deletions and amplifications.
Because linearity and genetic instability are two of the principal features of Streptomyces chromosomes, it was tempting to explain the second by the first. However, circularized chromosomes are even more unstable than linear ones. This instability of circular chromosomes could result from the lack of replication terminators and other functions that have evolved in bacteria with circular chromosomes to ensure a stable propagation of genetic information. Why has Streptomyces developed linear chromosomes instead of circular ones as in most bacteria? There are several possible reasons. Interactions between chromosomes are probably frequent in the Streptomyces multinucleoidal mycelium. Recombination by single cross-overs between circular chromosomes would lead to multimers, which would have to be resolved again to ensure proper segregation, whereas linear chromosomes can undergo homologous recombination events without generating multimers. Linear chromosomes may also allow for easier interspecies gene transfer, because single cross-overs between linear chromosomes and conjugative linear plasmids can exchange large segments of genetic material, which can then be transferred at high frequency into other strains. Such an event was postulated for the oxytetracycline biosynthesis gene cluster of S. rimosus (Gravius et al., 1994). There are also earlier reports of high-frequency transfer of unstable markers, such as antibiotic biosynthesis or resistance genes and AUDs (Baltz and Seno, 1988), which might have happened by similar events. Similarly, the mobilization of whole chromosomes into related strains would allow the exchange of chromosomal arms by single cross-over events at conserved sequences. Such recombinants might survive if the order of essential genes is also conserved.
Finally, deletions and amplifications can lead to changes in gene expression, reorganization of gene arrangements and probably even the creation of new coding sequences in both linear and circular chromosomes. Very unstable circular chromosomes carrying rearrangements are not necessarily destined for destruction by accelerated deletion. They could regain the less unstable linear structure by a single cross-over with the linear plasmids present in many streptomycetes, thus saving the rearrangements they underwent. Although most rearrangements are likely to be neutral or deleterious, genetic instability has an enormous potential to create the huge diversity of strains and secondary metabolites seen today and might be an important strategy in responding to the environmental challenges in nature.
We thank David A. Hopwood for critical reading of the manuscript. J.-N. Volff was supported by the European Molecular Biology Organization (EMBO).