Streptomyces genomes: Circular genetic maps from the linear chromosomes

¹ Present address: Department of Family Medicine, Taipei Veterans General Hospital, Shih-Pai, Taipei, Taiwan

² Present address: Food Industrial Research Institute, Hsin-Chu, Taiwan

*Corresponding author:

Carton W. Chen

Institute of Genetics, National Yang-Ming University

Shih-Pai, Taipei 112, Taiwan

Phone: (+886)-2-2826-7040

Fax: (+886)-2-2826-4930

Email: cwchen@ym.edu.tw

SUMMARY

The Streptomyces chromosomes are linear DNA molecules, and yet their genetic maps based on linkage analysis are circular. The only other known case like this has been the bacteriophage T2/T4, whose linear genomic sequences are circularly permuted and terminally redundant, and whose replication intermediates include long concatemers. These structural and functional features that contribute to the circularity of genetic maps are not found in Streptomyces. Instead, the circularity of the Streptomyces genetic maps appears to be caused by a completely different mechanism postulated by Stahl & Steinberg (Genetics 50: 531-538, 1964): a strong bias toward even numbers of crossover during recombination that creates misleading genetic linkages between markers on the opposite arms of the chromosome. This was demonstrated in physical inspection of the telomeres in recombinant chromosomes in interspecies conjugation promoted by a linear or circular plasmid. The preference for even numbers of crossovers is probably demanded by the merozygosity of the recombining chromosomes, and by the association between the telomeres mediated by interactions of covalently bound terminal proteins.

INTRODUCTION

An emerging enigma from the discovery of the linear chromosomes of the filamentous bacteria Streptomyces (Lin et al., 1993) is the circularity of their genetic maps (reviewed in Hopwood, 1967).

The circularity of the genetic maps was first established in Streptomyces coelicolor A3(2) through extensive conjugation analysis (Hopwood, 1965; Hopwood, 1966). Subsequently, it was extended to other species (for example, Friend & Hopwood, 1971) and to protoplast fusion analysis. All the linkage analyses in different Streptomyces species have been consistent with the circularity of the genetic maps without exceptions.

In his papers, Hopwood gave these discerning remarks: 'while the linkage map of S. coelicolor is circular, it appears at present to be impossible to decide by this kind of formal analysis whether the chromosome is circular or linear' (Hopwood, 1965), and 'whether, like that of E. coli, the genome of S. coelicolor can actually be a closed loop, remains to be determined' (Hopwood, 1966). These statements were based on the careful conclusion of 'lack of constant genome ends' in S. coelicolor, which did not rule out the possibility that the S. coelicolor chromosome is a linear DNA with circular permuted sequences and variable ends as in the closely related T2 and T4 phages (Streisinger et al., 1964).

The T2/T4 model has been the only other known case for a linear replicon to exhibit a circular genetic map. The sequences of these linear genomes are circularly permuted and contain terminal direct repeats, which were generated by formation of linear concatemers during replication followed by headful packaging of more than one genome equivalent of DNA. The circular permutation of the T2 and T4 DNA sequences means that the linkage of any pair of proximal markers on the circular genetic map is interrupted only in rare molecules, where they occupy opposite ends of the DNA. This 'lack of constant genomic ends' (Hopwood, 1966) thus leads to a circular genetic map.

The T2/T4 prototype is, however, not applicable to the Streptomyces chromosomes, which, although are also linear, turned out to have very different basic structures. The chromosomal DNA of Streptomyces (mostly about 8 Mb) contain proteins covalently bound at the 5' ends (Lin, et al., 1993). Their overall sequences are not circularly permuted as in T2/T4 (Leblond et al., 1993; Lin, et al., 1993). The chromosomal ends are fixed, and contain long (24- to 500-kb) inverted repeats (Leblond, et al., 1993; Lezhava et al., 1995; Lin, et al., 1993; Pandza et al., 1997; Redenbach et al., 1993).

A completely different model was proposed more than thirty years ago by Stahl & Steinberg (1964), and was also taken into account by Hopwood (1965; 1966; Stahl, 1967) in his linkage analysis of S. coelicolor. In this hypothetical scenario, the linear replicons recombine with a bias toward even numbers of crossovers, such that markers at opposite ends of each mating molecule tend to finish in the same recombinant molecule (instead of segregating independently to different ones). The result of this is that the terminal markers would exhibit genetic linkage, which would then close the genetic map into a circle.

There have been no hints whether this scenario may be applied to Streptomyces (Hopwood, 1965; Hopwood, 1966; Stahl, 1967). Little is known about the biochemical mechanisms of conjugal transfers and recombination of these linear chromosomes. Conjugation in Streptomyces occurs on surface culture, when two mating cultures are mixed and grown together . Conjugation is mediated by naturally occurring linear or circular plasmids, both of which are abundant in Streptomyces. Some naturally integrated plasmids also appear to be conjugative, such as the integrated plasmid SLP1 in S. coelicolor (Bibb et al., 1981), which on transfer to S. lividans can be circularised or integrated into the chromosome via a site-specific recombination system.

The conjugative plasmids are transferred from the donor mycelium to recipient mycelia (intermycelial transfer), and presumably spread along the recipient mycelia (intramycelial transfer; Kieser et al., 1982). The transfer of chromosomes during Streptomyces conjugation cannot be studied biochemically, and can be detected only by the appearance of recombinants. The transfer of the Streptomyces chromosomes has been assumed (without proof) to be essentially the same as that of the conjugative plasmid: from donor to recipient, although there are circumstances in which back transfers of the chromosome from the recipient to the donor were also indicated (Hopwood, 1984).

Recombination has been postulated to occur mostly in the merozygote (partial diploid) state during conjugation in most bacteria studied including Streptomyces (Hopwood, 1967), but probably not in protoplast fusion in which complete genomes are expected to be involved in recombination (Hopwood & Wright, 1978). Studies of recombination during protoplast fusion in Streptomyces have also been consistent with circular genetic maps.

In this study, taking advantage of the possibility to distinguish between the termini of the chromosomes of S. coelicolor and S. lividans by hybridisation (Huang et al., 1998), we examined the terminal sequences of the recombinant chromosomes resulting from interspecies mating, from which we inferred the numbers of crossovers that had taken place between the parental chromosomes. An odd number of crossovers would result in a recombinant chromosome with one telomere from each parental chromosome, whereas an even number of crossovers would lead to a recombinant chromosome with both telomeres from the same parent. Here we present the physical evidence, which supports the Stahl & Steinberg model, i.e., the circular genetic maps of the Streptomyces are generated by a strong bias for even numbers of crossovers.

METHODS

Bacterial strains and plasmids. S. coelicolor and S. lividans strains used are listed in Table 1 with their respective genetic markers and plasmid status. The Streptomyces cultures were maintained on YEME (Hopwood et al., 1985) agar at 30 ûC.

Conjugation analysis. Conjugation between S. coelicolor and S. lividans was performed according to Hopwood et al. (1985). About 10⁷ - 10⁸ spores of each parental culture were mixed and plated on ISP medium 2 (ATCC culture medium 196; yeast extract 4 g, malt extract 10 g, dextrose 4 g, agar 20 g per liter, pH 7.3). Spores were collected after incubation at 30 ûC for about 7 days, and recombinant cultures were isolated on minimum medium containing the appropriate supplements. Each parental culture alone was handled the same fashion to obtain the background frequency of spontaneous mutation giving rise to the selected recombinant phenotype. Agarase activities were revealed by direct visualisation of the degradation of agar surrounding the colonies.

Physical and chemical analyses of Streptomyces chromosomes.

For BamHI restriction analysis, chromosomal DNA was isolated, digested, and subjected to Southern hybridisation according to Hopwood et al. (1985). The hybridisation probes were the terminal 1.4-kb BamHI fragment of the S. coelicolor chromosome and the terminal 0.65-kb BamHI fragment of the S. lividans chromosome (Huang, et al., 1998). For AseI restriction analysis, preparation, PFGE, and Southern hybridisation of the genomic DNA were as described previously (Lin, et al., 1993).

RESULTS

The chromosomes of the strains used in heteroclone analysis are linear

One piece of conclusive evidence for connecting the (three) linkage groups in S. coelicolor into a circle was from the analysis of 'heteroclones' (Hopwood, 1966), which were found among recombinant-type colonies when a pair of very close markers was selected. The heteroclones are unstable merodiploids that, on subsequent subculturing, segregate into parental and various recombinant genotypes, which are convenient subjects for conventional linkage analysis. The results of the analysis distinctly connected the previously separate linkage groups, including the two at the opposite arms of the S. coelicolor A3(2) chromosome, into a circle (Hopwood, 1966).

Because Streptomyces cultures occasionally undergo spontaneous deletions, leading to circularisation of the chromosomes (Lin, et al., 1993), there exists a possibility that the mutant strains used in the heteroclone experiments had circular chromosomes, which would account for the final circular linkage map. To examine this possibility, the chromosomal DNA of two parental strains studied by Hopwood (1966), 773 and 928 (D. A. Hopwood, personal communication), was isolated and subjected to restriction and PFGE analysis. The AseI restriction patterns (not shown) of these strains were identical to that of S. coelicolor M145 except for a fragment that contained the integrated SCP1 (Kieser et al., 1992). The terminal sequences of these two chromosomes were shown to be intact by hybridisation analysis using the telomere DNA of S. coelicolor chromosome as the probe (not shown). These results indicated that the 773 and 928 chromosomes were linear, and thus circularity of the parental chromosomes in the heteroclone experiment is not a possible explanation for the 'lack of constant genomic ends' (Hopwood, 1966).

The basic strategy of crossover analysis

To test Stahl & Steinberg's model for the generation of circular genetic maps of Streptomyces, one would need to differentiate between odd and even numbers of crossovers that have occurred between the linear chromosomes. A simple strategy is to isolate recombinant chromosomes and examine their ends. The presence of both ends from the same parent would indicate an even number of crossovers, whereas the presence of ends from different parents (mixed ends) would indicate an odd number of crossovers.

There are two practical prerequisites in the application of this strategy. Firstly, in order to reach a definite conclusion, ideally one must examine the very ends of chromosomes, not merely certain genetic markers located near the termini. Secondly, it is necessary to have a tool to inspect and distinguish the parental chromosomal ends experimentally. For recombination between two homologous chromosomes, such as intraspecies conjugation of Streptomyces, it is generally impossible to distinguish the ends of the two mating chromosomes, which are usually identical.

We have recently cloned terminal DNA from the chromosomes of several species of Streptomyces (Huang, et al., 1998). These terminal sequences, being highly conserved for only the first 166-168 bp, could be discriminated by restriction and hybridisation analysis. This thus makes it plausible to perform interspecies conjugation and examine directly the inheritance of the different parental telomeres in the recombinants.

For interspecies conjugation analysis we chose two model species, Streptomyces lividans 66 and Streptomyces coelicolor A3(2), of which many strains with defined genetic markers and plasmid status (circular or linear plasmid) were available. The other important consideration was the relative global arrangements of essential and housekeeping genes in the mating pairs. Gross deviations in the gene orders would demand complex or even illegitimate recombination schemes to achieve viable recombinant chromosomes with the selected phenotypes. The general orders of the genes that have been characterised in S. coelicolor and S. lividans are in relatively good correspondence (Hopwood, et al., 1985; Hopwood et al., 1983; Leblond, et al., 1993). The overall similarity in chromosomal sequences in the two species were further demonstrated by the similarity in their macro-restriction maps, in which most of the restriction sites (AseI and DraI) not only are similarly located [Fig. 3(c)], but also contain homologous neighbouring sequences as revealed by cross-hybridisation (Leblond, et al., 1993).

Wild type S. coelicolor A3(2) contains two free plasmids: the linear 350-kb SCP1 plasmid (Kinashi & Shimaji-Murayama, 1991) and the circular 31-kb SCP2 plasmid (Bibb & Hopwood, 1981), and an integrated SLP1 plasmid, which is occasionally excised and becomes a circular form on transfer to other species (Bibb, et al., 1981). Wild type S. lividans 1326 contains a free 50-kb linear plasmid SLP2 (Chen et al., 1993; Hopwood, et al., 1983). The right 15 kb of this plasmid is homologous to the first 15 kb at both ends of the S. lividans chromosomes (Chen, et al., 1993). Another conjugative plasmid SLP3, whose presence in S. lividans has only been determined genetically (Hopwood, et al., 1983), was absent from all the strains used in this study.

Hybridisation probes for the respective telomeres were the 1.4-kb BamH1 terminal fragment of the S. coelicolor chromosome and the 0.65-kb BamH1 terminal fragment of the S. lividans chromosome (Huang, et al., 1998). Each probe hybridised to its own chromosomal DNA, but not to the other under our conditions (Figs. 1, 2). The S. lividans probe also hybridised to the right end of SLP2 plasmid.

In this study, one of the mating pair ('donor') contained a free conjugative plasmid, while the other ('recipient') did not possess any free plasmid. Thus, 'donors' and 'recipients' are defined with respect to the free conjugative plasmids, the direction of transfer of which is unambiguous. The transfer of the chromosomes is assumed argumentatively to be identical to that of the plasmid, but may not be the case (see below).

The integrated plasmids contributed little, if any, to interspecies recombination

No recombination was detectable in conjugation between SLP2^- SLP3^- strains of S. lividans (Hopwood, et al., 1983). On the other hand, in mating between the SCP1^- SCP2^- strains of S. coelicolor the chromosomes recombine at a low but real frequency of about 10^-8, presumably mediated by the integrated SLP1 plasmid present in all S. coelicolor A3(2) strains (Bibb and Hopwood, 1981). We set out to determine the extent that the integrated plasmid SLP1 (and any other) might contribute to chromosome recombination in our interspecies situation. A cross was performed between two strains devoid of free plasmids, S. lividans ZX7 (pro-2 str-6 rec-46 DdndA SLP2^- SLP3^-) and S. coelicolor M145 (SCP1^- SCP2^-; Hopwood, et al., 1985). The DdndA mutation in ZX7 removed a DNA modification system that causes DNA degradation during PFGE (Zhou et al., 1988), and the rec-46 mutation decreased plasmid recombination (Tsai & Chen, 1987), but not chromosome recombination during conjugation (Kieser et al., 1989). M145 was a prototroph cured of two intrinsic plasmids SCP1 and SCP2, but still possessed the integrated plasmid SLP1 (Bibb, et al., 1981).

Proline prototrophic (Pro⁺) and streptomycin resistant (Str^r) cultures were isolated among the spores from the mixed growth of the two parents (Hopwood, et al., 1985) at a frequency of about 8.0 x 10^-8. This was not significantly different from the spontaneous occurrence of the same phenotype in M145 alone (4.3 x 10^-7; presumably through a mutation to streptomycin resistance) or in ZX7 alone (5.4 x 10^-8; presumably through a reversion to proline prototrophy). Therefore, the contribution of SLP1 to chromosomal recombination in our interspecies mating system is negligible, if any, and may be safely ignored in the following analyses.

Chromosomes recombine with a bias toward even numbers of crossovers in linear plasmid-mediated conjugation

We next analyzed conjugation mediated by the linear plasmid SCP1. Mating was performed between S. coelicolor M146 (hisA1 uraA1 strA1 SCP1+ SCP2-; Hopwood, et al., 1985), and plasmidless S. lividans ZX7. His⁺ Pro⁺ recombinants were isolated at a frequency of 1.5 x 10^-6, comparable to SCP1-mediated intraspecies recombination in S. lividans (3.5 - 6.8 x 10-6; Hopwood, et al., 1983) or S. coelicolor (about 10-6; Bibb and Hopwood, 1981). In contrast, spontaneous mutations giving rise to the selected His⁺ Pro⁺ phenotype appeared at much lower frequencies in M146 (5.3 x 10^-8) or ZX7 (9.2 x 10^-8). Analysis of the chromosomal DNA from 10 recombinants by hybridisation revealed that all had inherited both telomeres from M146 [Fig. 1(a)]. No recombinant chromosomes with mixed ends were observed. This result indicated that an even number of crossovers had occurred in these recombinants, while only one crossover was necessary to generate His⁺ Pro⁺ recombinants, an additional crossover must have occurred between the hisA1 marker and the left telomere. In theory, the extra crossover might happen between pro-2 and the right telomere, giving rise to recombinant chromosomes with both ends from ZX7 instead. Interestingly, these alternative recombinant products were not found.

Further inspection of the non-selected phenotypes in these His⁺ Pro⁺ recombinants revealed that all of them (15/15) were uracil auxotrophic and agarase positive, both of which were present only in S. coelicolor M146 (S. lividans does not possess the agarase gene).

A similar result was obtained using another linear plasmid SLP2. Conjugation between S. lividans ZX7 [SLP2] and S. coelicolor M145 yielded Pro⁺ streptomycin-resistant (Str^r) recombinants at a frequency of 2.7 x 10^-5, comparable to SLP2-mediated intraspecies recombination in S. lividans (1.7 - 3.2 x 10-4; Hopwood, et al., 1983) or S. coelicolor (5 x 10-5; Bibb and Hopwood, 1981), and much higher than the frequencies of 1.0 x 10^-7 and 4.3 x 10^-7, respectively, for spontaneous appearance of Pro⁺ Str^r in ZX7 [SLP2] and M145. DNA from 11 of these recombinants were isolated and hybridised with the terminal probes [Fig. 1(b)]. No hybridisation was observed with terminal probe of S. coelicolor, while the terminal probe of S. lividans hybridised to all recombinant chromosomes. In 10 out of the 11 samples, it hybridised indistinguishably to the terminal DNA of the recombinant chromosomes and the right end of SLP2 plasmid. In the other (No.4), there was a weaker hybridisation to a 10-kb BamHI fragment. This indicated that the chromosomal termini in this strain had undergone deletions (probably resulting in circularization; Lin, et al., 1993). The chromosomal deletions might have caused the loss of SLP2 (Denapaite & Cullum, 1998; Lin, 1998; Wu, 1994).

In another cross between S. lividans ZX7 [SLP2] and plasmidless S. coelicolor M130 (hisA1 uraA1 strA1 SCP1^- SCP2^-), Pro⁺ Ura⁺ recombinants were isolated at a frequency of 1.0 x 10^-6 against the background mutation frequencies of about 10^-8 in the parents. Chromosomal DNA from ten recombinants analyzed also contained only the telomeres of the donor S. lividans (data not shown). Thus, the observed non-reciprocity in recombination did not appear to be limited to a particular selection scheme.

Examination of the unselected markers in these two crosses (ZX7 [SLP2] x M145 and ZX7 [SLP2] x M130) revealed that the majority (9/12) of them were of the donor type, and the remaining (3/12) were of a mixed type (agarase positive and His^-).

Chromosomes recombine with a bias toward even numbers of crossovers in circular plasmid-mediated conjugation

The preference for even numbers of crossovers was also observed in conjugation promoted by a circular plasmid. In crosses between S. coelicolor 1190 (hisA1 uraA1 strA1 SCP2⁺) and plasmidless S. lividans ZX7, His⁺ Pro⁺ recombinants were isolated at a frequency of 1.1 x 10^-6 against spontaneous mutation frequencies of 3 - 9 x 10^-8 in the parent stains. SCP2-mediated intraspecies recombination occurs at about the same frequencies (10^-6) in S. coelicolor (Bibb and Hopwood, 1981). Chromosomal DNA from 12 recombinants was isolated and examined: 10 of them contained only the telomeres of ZX7, while the other 2 (Nos. 5 and 9) contained mixed telomeres: one from each parent [Fig. 2(a)]. The two recombinant cultures with mixed ends grew normally, and the observed mixed telomeres persisted on extended subculturing and in each reisolated culture [Fig. 2(b)]. This indicated (i) that the observed mixed ends existed on the same chromosome, rather than reflecting the presence of a mixed population of two chromosomes each with the same telomeres from different parents, and (ii) that the recombinant chromosomes with the mixed telomeres did not suffer from obvious growth disadvantage.

All the non-selected genetic markers in these recombinants were those of the recipient ZX7 (Ura⁺ and agarase negative). While a single crossover (between hisA1 and pro-2) was sufficient to give rise to the selected His⁺ Pro⁺ phenotype, for the 10 recombinants with both recipient telomeres an additional crossover must have occurred between pro-2 and dag. For the two recombinants with mixed ends, at least one (or a larger odd number of) crossover must have occurred either to the left of hisA1 or to the right of uraA1. The AseI restriction analysis (see later) indicated that the former had taken place.

In the repeat of the same cross, Pro⁺ Ura⁺ recombinants were isolated at a frequency of 2.1 x 10^-6. Eighteen recombinants were picked and their chromosomes examined. In all of them, only the termini of the S. lividans chromosome were present (data not shown). All the non-selected genetic markers are those of the recipient ZX7 (His⁺ and agarase negative). No recombinants with mixed ends were found.

From these results we may conclude that (i) there was a strong bias for even numbers of crossovers between the recombining chromosomes during conjugation, and (ii) there was a surprising correlation between the topology of the conjugative plasmids and the parental telomeres inherited by the recombinants: the telomeres and genetic markers from the plasmid donor were predominantly inherited in linear plasmid-driven conjugation; whereas the telomeres and genetic markers from the recipients were preferentially maintained in circular plasmid-driven transfer.

PFGE analysis of the recombinant chromosomes

The recombinant chromosomes of two crosses were examined by AseI digestion and PFGE (Fig. 3). The restriction patterns of these chromosomal DNA molecules confirmed the foregoing conclusions regarding the inheritance patterns of the recombination chromosomes. The origins of many fragments in the recombinants could be readily assigned to one or the other parent. No obviously new fragments resulting from recombination were present, indicating that relatively few crossovers had taken place.

Most of the AseI fragments in the 4 His⁺ Pro⁺ recombinant s in the M146 [SCP1] x ZX7 cross appeared to originate from the donor M146 [Fig. 3(a)]. While several fragments from the two parents could not be readily distinguished, all those that could corresponded to the fragments of S. coelicolor, such as fragments C, D, E, I, J, L, and K (Kieser, et al., 1992). In contrast, fragments B, C, E₁, E₂, and G of S. lividans (Leblond, et al., 1993) were absent. Tracing the organisation of the recombinant chromosomes according to the genetic markers and the restriction fragments revealed that it could be achieved by a minimum of two crossovers: one between hisA1 and pro-2, and the other close to the left of his-2 [panel (c)]. Although the two events supposedly occurred on fragments C (and probably I) of S. coelicolor, the sizes of these two fragments appeared unchanged, probably because the corresponding restriction sites in the S. lividans chromosome were similarly located. Overall, the prevalence of the donor sequences in these recombinants was apparent.

In the 1190 [SCP2] x ZX7 cross, both the His⁺ Pro⁺ and Ura⁺ Pro⁺ recombinants were analyzed [Fig. 3(b), (d), (e), (f)]. The three His⁺ Pro⁺ recombinant chromosomes (Nos. 3, 4, 7) with both recipient ends were almost in perfect contrast to their counterparts in the M146 [SCP1] x ZX7 cross [panels (a), (c)]: The S. coelicolor fragments D, E, I, J, and L were absent, and the S. lividans fragments C, E₁, E₂, E₃, F, H₁, andH₂ were present [panel (b)]. This pattern suggested that, in addition to the crossover between hisA1 and pro-2 (on fragment C of S. coelicolor and fragment B of S. lividans), another one occurred between pro-2 and dag [panel (d)]. This placed the dag⁰ and uraA1⁺ markers and the S. lividans telomeres on the recombinant chromosomes. The terminal probe of S. lividans hybridised to the terminal H₁fragment of S. lividans and one of the largest fragments (presumably A fragment of S. lividans; data not shown), supporting the deduction and the results of the BamHI analysis (Fig. 2). The presence of the H fragment of S. coelicolor and the absence of the G fragment of S. lividans (both of which contained the uraA1 locus) suggested the possible occurrence of an additional double crossover in this region [panel (d)].

In the two His⁺ Pro⁺ recombinant chromosomes with mixed ends (Nos. 5, 9), both parents appeared to contribute multiple fragments, such as fragments H, I, J, and L from S. coelicolor, and fragments C and F from S. lividans [panels (b), (e)]. Most significant is the presence of the J fragment at the left end of the S. coelicolor chromosome. This fragment hybridised to the terminal probe of S. coelicolor. The presence of the S. lividans end on the A fragment (right end of chromosome) was also demonstrated by hybridisation to the S. lividans terminal probe (not shown). The H₁ fragment containing the left end of the S. lividans chromosome was absent, based on the lack of hybridisation of the S. lividans terminal probe to the DNA corresponding to the co-migrating H₁ andH₂ fragments (not shown), and the diminished fluorescent intensity of this population [panel (b)]. These results, consistent with the mixed ends observed among the BamHI fragments (Fig. 1), indicated one (or another odd number) additional crossover to the left of the hisA1 marker in these two recombinants [Fig. 2(a)]. Like the recombinant chromosomes with the same ends, these recombinant chromosomes appeared to contain the H fragment of S. coelicolor but not the G fragment of S. lividans, suggesting a possible double crossover in this region.

The Ura⁺ Pro⁺ recombinants of the 1190 [SCP2] x ZX7 cross exhibited a very similar to those of the His⁺ Pro⁺ recombinants. Compared to the His⁺ Pro⁺ recombinants in the same cross, the most notable difference was the preservation of the G fragment of S. lividans and missing of the H fragment of S. coelicolor. The reason for the difference is not clear. It may presumably reflect the difference in the selection schemes. The terminal probe of S. lividans hybridised to fragments A and H₁, whereas no hybridisation was observed by the terminal probe of S. coelicolor (not shown), consistent with the previous conclusion based on BamHI analysis [Fig. 2(a)].

DISCUSSION

This study provides the first experimental demonstration of Stahl & Steinberg's hypothetical model (1964) for generating circular genetic maps operationally from linear replicons by a constraint in the recombination process. The biased production of recombinant chromosomes with even-numbered crossovers, predicted in this model, was demonstrated in interspecies conjugation of Streptomyces mediated by either a linear or a circular plasmid.

The even numbered crossovers, however, produced only one of the two types of products expected from reciprocal recombination, and the type of recombination products recovered appeared to depend on the topology of the conjugative plasmids. In circular plasmid-mediated conjugation (with only one plasmid tested), most of the recombinant chromosomes possessed the recipient-type termini. The reciprocal products, i.e., recombinant chromosomes with the donor-type ends, were not recovered. The preponderance of unselected markers as well as the AseI restriction fragments of the recipient type in these recombinants suggested that the recipient chromosomes had picked up only a relatively small internal segment(s) of the donor chromosome that was presumably mobilised from the donor cell. This merozygous state of recombination resembles that in E. coli conjugation (also mediated by circular plasmids), in which the donor chromosome rarely enters the recipient cell in its entirety. Bibb & Hopwood (1981) did not observed this strongly preferred inheritance of recipient markers in intraspecies mating between SCP2⁺ and SCP2^- S. coelicolor. It has been proposed that the inheritance of donor chromosomal markers depends on recombination in the immediate recipient 'cells' as well as subsequent recombination in adjacent cells on further (intramycelial) transfer of the recombinant chromosomes (Lydiate et al., 1985). The partial donor chromosome in the recipient is not or rarely transferred intramycelially unless incorporated into an intact chromosome through homologous recombination. Thus, the shortage of donor markers in the recombinant chromosomes in the interspecies SCP2⁺ x SCP2^- crosses may simply reflect the lower efficiency of recombination between the chromosomal sequences of S. coelicolor and S. lividans due to lower sequence homology.

In contrast, in the linear plasmid-mediated conjugation of Streptomyces (with two plasmids tested), double crossovers gave rise to recombinant chromosomes consisting mainly of donor chromosome plus an internal stretch(es) of the recipient chromosome. Again, the reciprocal products, i.e., recombinant chromosomes with both ends of the recipient-type, were not found. This predominant presence of donor markers in the recombinants has also been observed in the classical SCP1⁺ x SCP1^- crosses in S. coelicolor by Hopwood et al. (1973). Although the SCP2 status in this earlier study was not yet established, the absence of the tendency for progeny to inherit donor markers in SCP1^- SCP2⁺ x SCP1^- SCP2^- cross of S. coelicolor (Hopwood, 1984) indicated that the linear plasmid SCP1 was involved in the skewed inheritance pattern. These authors proposed that some high-frequency donors (like Hfr strains of E. coli) present in SCP1⁺ culture contributed markers preferentially to the progeny. An alternative conjecture is presented below.

The selectivity in recombinant types recovered was unlikely due to the inhibition exerted by one particular parental species to the other during conjugation. Although SCP1 encodes an antibiotic methylenomycin A (Wright & Hopwood, 1976), which may inhibit SCP1^- strains, there appears to be no or little effect of the methylenomycin A on the patterns of chromosome recombination during conjugation or protoplast fusion (Hopwood, 1984; Kirby & Hopwood, 1977). Furthermore, the other linear plasmid SLP2, which does not encode any detectable antibiotic activities, display the same skewed inheritance. Thus, the plasmid status (donor/recipient), rather than species, appeared to determine the inheritance of non-selected sequences (including the telomeres) on the recombinant chromosomes selected.

While the contrasting effect exerted by plasmid topology on chromosome inheritance is intriguing, the very limited number of plasmids tested (one circular and two linear) may not justify formulating a general rule. Further studies employing more plasmids of both categories are necessary for validation. Nevertheless, whether the postulation is confirmed or not, it does not affect the support that the current study lends to the hypothetical scenario postulated by Stahl & Steinberg (1964): a circular genetic map generated by preferred even numbers of crossovers between linear genomes.

If the preference for even number of crossovers is real, what is the biochemical basis for this constraint? An even number of crossovers is necessary for linear chromosomes, if recombination occurs in merozygotes consisting of an intact chromosome and an internal chromosomal sequence [Fig. 4(b)]. An odd number of crossovers would result in two partial chromosomes each lacking a telomere. This scenario probably represents the recombination events in circular plasmid-mediated conjugation of Streptomyces. The partial exogenote is likely mobilised from the donor by the circular plasmid from an internal origin on the chromosome. Occasionally, the transfer may reach one end of the donor chromosome, which would make odd-numbered crossovers possible [Fig. 4(a)], resulting in a recombinant chromosome with mixed ends.

The same merozygote situation may also apply to chromosome recombination in linear plasmid-mediated conjugation, but the roles of the mating partners would have to be reversed to be consistent with the absence of recombinant chromosomes with both recipient ends: The intact chromosome would have to be that of the donor, and the fragment would be of the recipient. On the other hand, if we assume that this recombination step takes place in the recipient (of plasmid), this would imply that the chromosomes are transferred in total (or nearly so) from the plasmid donor, and this was demonstrated in the restriction analysis (Fig. 3). The complete transfer of donor chromosomes in linear plasmid-mediated conjugation agrees with the model proposed by Chen (1996), in which the chromosome transfer mediated by linear plasmids starts from the ends. The principle underlying the unilateral disappearance of the recipient chromosomal sequences can only be speculated on at present, but it suggests the devastation of the recipient chromosomes by a selfish and invasive action of the donor chromosomes. The alternative scenario, in which mainly partial recipient chromosome is transferred into the donor cell (in opposite direction as the plasmid transfer), would also be a remarkable and uncommon phenomenon.

Although protoplast fusion was not used directly to address the topology of the genetic maps of Streptomyces, a constraint for even numbers of crossovers has always been imposed successfully in the formal analysis of the genetic linkage (For example, Hopwood and Wright, 1978). The rationale is based on the assumption that intact circular chromosomes are involved in recombination, and an odd-number of crossovers would result in dimeric chromosomes. Interestingly, the circular genetic maps based on this imposed condition have been essentially consistent for S. coelicolor and other species. Now that the assumption of circular chromosomes was incorrect, there must be a different reason for the demand for even-numbered crossovers. Whether the mating linear chromosomes are intact or not in protoplast fusion, there is no a priori topological constraint that demands even numbers of crossovers [Fig. 4(c)]. Odd-numbered crossovers give rise to a recombinant chromosome with mixed ends, whereas even-numbered crossovers, same ends. An even number of crossovers would be necessary, if the ends of the chromosomes are not free [Fig. 4(d)]. This is true even if the chromosomes are fragmented as long as complete genomes are to be restored. A likely constraint is a strong interactions between the telomeres, mediated by the covalently bound terminal proteins (TP), as having been observed in the other linear replicons with TP, adenoviruses (Robinson et al., 1973) and f29 (Ortin et al., 1971). Circular configurations of these linear viral DNA molecules were observed, when released from the virion particles without proteolytic or detergent treatment.

Moreover, during electrophoresis, TP-capped DNA of these linear replicons also appears to form a proteinase- and detergent-sensitive complex, which cannot enter the gel. The linear chromosomes and plasmids of Streptomyces also exhibit the same behavior, suggesting strong interactions between the TP (Lin, et al., 1993), which may be implicated in important biological functions, such as replication, conjugal transfer, and structural stability (Chen, 1996).

The strength of the constraint imposed by the terminal interactions on the chromosome recombination should depend on the strength of the terminal interactions. The existence of rare recombinants with mixed ends indicates that the terminal interactions are not perfect during recombination and exchange of telomeres is still possible.

The scheme portrayed in Fig. 4(c) implies that both products of reciprocal recombination, i.e., recombinant chromosomes with both ends from either parent, may be recovered from protoplast fusion. This hypothesis remains to be tested experimentally. Since plasmids are not necessary for recombination in protoplast fusion, it is interesting to investigate their effect on the outcome of the recombination, if any.

An alternative to the imposed even-numbered crossovers scenario is that there is no constraint on the numbers of crossovers allowed, but that the products of even-numbered crossovers are selectively recovered. This would imply that the products of odd-numbered crossovers, i.e., chromosomes with mixed ends, are biologically disadvantageous. The rarity of recombinant chromosomes with mixed ends suggests that the selection against them would be severe in this hypothesis. Yet the stable maintenance of the S. coelicolor-S. lividans mixed ends in the recombinant cultures (Nos. 5 and 9) and their normal growth characteristics do not support this view.

A third theoretical possibility exists that is independent of the numbers of crossovers: It is possible that identical telomeres are preserved on the recombinant chromosomes through conversion of one terminal sequence into the other. Again, the stable maintenance of mixed ends in the recombinant chromosomes on prolonged subculturing without signs of conversion reduces the likelihood of conversions. Moreover, if conversion is involved, it must exert a specific directionality in a very selective way, i.e., donor to recipient conversion of telomere sequences in circular plasmid-mediated conjugation, and conversion in the opposite direction in linear plasmid-mediated conjugation. The range spanned by this homogenotisation would also have to be relatively long: at least 1.4 kb (to reach the restriction sites used in the analysis). Qin & Cohen (1998) in their study of the replication of the linear plasmid pSLA2 of Streptomyces rochei failed to observe any homogenotisation between the two termini.

Of the three possible hypotheses: (i) even-numbered crossovers constrained demanded by merozygosity and by terminal interactions, (ii) selective recovery of even-numbered crossovers, and (iii) homogenotisation; the first one, with its in vitro evidence and interesting biological implications, is currently the working model of our choice.

While our systematic analysis was performed using only one free conjugative plasmid in each cross, the classical conjugation studies of Streptomyces are often performed with the involvement of more than one free plasmid (circular and linear) and sometimes in both parental strains. The directionality of transfer (of the plasmids and of the chromosomes) would be more difficult to predict or establish, and the molecular genetic analysis would be more complex. However, there are no obvious doubts that the constraints for even-numbered crossovers would be absent in multi-plasmid schemes.

In a broader scope, the materialisation of the Stahl & Steinberg model indicate that the danger of equating the topology of the genetic maps to that of the genomes is not limited to the special case of T2/T4 phages. The well-founded prudence displayed by Hopwood (1965; 1966) more than three decades ago should still be observed vigorously in the physical and sequence analyses of microbial genomes. For the latter, the popular shotgun sequencing approach in assembling whole genome sequences, if applying to T2/T4 genomes, would certainly give a circular physical map. Therefore, not only is the topology of circular genetic maps not a reliable guideline, the topology derived from assembled sequence contigs is also not free from similarly hidden pitfalls and traps.

ACKNOWLEDGEMENTS

We thank Nora S. Chen and Caroline Hu for technical assistance, Helen Kieser, Tobias Kieser, and David A. Hopwood for the Streptomyces strains, and D. A. Hopwood and Franklin W. Stahl for reading of the manuscript and suggestions for improvement. The examination of the cultures used in heteroclone analysis was suggested by D. A. Hopwood. This work is supported by a research grant from National Science Council (NSC87-2316-B010-M44) and from National Health Research Institute (DOH87-HR-713). C. W. C. was a recipient of a research award from the Medical Research and Advancement Foundation in memory of Chi-Shuen Tsou.

Table 1. Streptomyces strains used in this study

Strain	Genotype	Source/reference
S. coelicolor M130	hisA1 uraA1 strA1 SCP1- SCP2-	Hopwood, et al., 1985
M145	SCP1- SCP2-	Hopwood, et al., 1985
M146	hisA1 uraA1 strA1 SCP1+ SCP2-	Hopwood, et al., 1985
1190	hisA1 uraA1 strA1 SCP1- SCP2+	Hopwood, et al., 1985
S. lividans ZX7	pro-2 str-6 rec-46 DdndA SLP2- SLP3-	Zhou et al., 1988
ZX7 [SLP2]	ZX7 containing SLP2 plasmid	Chen et al. 1993

Legend to figures

Fig. 1. Inheritance of telomeres during interspecies conjugation mediated by linear plasmids. His⁺ Pro⁺ recombinants were isolated from conjugation between S. coelicolor M146 [SCP1] and S. lividans ZX7 (a), and Pro⁺ Str^r recombinants from conjugation between ZX7 [SLP2] and S. coelicolor M145 (b). Chromosomal DNA from these recombinant cultures was digested with BamHI and hybridised with the 1.4-kb BamHI terminal DNA of S. lividans chromosome and the 0.65-kb BamHI terminal DNA of S. coelicolor chromosome (shown in enlarged views). The parental chromosomes are represented by the horizontal lines showing respective genetic markers and TP (filled circles). The solid triangles depict the selected alleles, and the open triangles, the non-selected markers found in all or most of the recombinants; dag represents the agarase gene present in S. coelicolor, but not in S. lividans (dag⁰). The two minimum crossovers necessary for the observed recombination are indicated by the trace line between the two chromosomes. In this orientation, the terminal AseI J fragment of S. coelicolor and the terminal H₁ fragment of S. lividans are to the left.

Fig. 2. Inheritance of telomeres during interspecies conjugation mediated by circular plasmids. (a) His⁺ Pro⁺ recombinants were isolated from conjugation between S. coelicolor 1190 [SCP2] and ZX7, and their chromosomal DNA was analyzed as described in Figure 1. Two recombinants (5 and 9) harbouring mixed telomeres are shaded in black. The least addition crossover necessary for the appearance of the mixed ends in these two recombinants is indicated by the dashed line. (b) Chromosomal DNA from three reisolates each of recombinants 5 (strains 5-1 to 5-3) and 9 (stains 9-1 to 9-3) was purified and subjected to the same restriction and hybridisation analysis.

Fig. 3. Restriction patterns of the recombinant chromosomes. The genomic DNA from selected recombinants from the M146 [SCP1] x ZX7 cross (a), and the 1190 [SCP2] x ZX7 cross (b) was digested by AseI and separated by PFGE. The designations of the AseI fragments of the parental chromosomes are indicated on both sides. The two recombinants (5 and 9) with mixed ends are designated by the shaded numbers. In panel (a), S, S1 and S2 indicate the intact SCP1 (350 kb), and its two larger (180- and 140-kb) AseI fragments, respectively (Kinashi and Shimaji-Murayama, 1991). Of these, S2 appeared to co-migrate with the M fragment of the S. coelicolor chromosome (Kieser, et al., 1992), resulting in the elevated fluorescent intensities. The smallest (18-kb) AseI fragment of SCP1 might also co-migrate with the O fragment of the chromosome. In panel (b), the presence of mixed ends in recombinants Nos. 5 and 9 was demonstrated by the presence of the fragment corresponding to the J fragment of S. coelicolor, which hybridised to the S. coelicolor end probe, and the largest fragment corresponding to the A fragment of S. lividans, which hybridised to the S. lividans end probe (hybridisation data not shown). The larger fragments in the same cultures were underrepresented in these experiments probably due to breakdown and/or entrapment. Panels (c) - (f) depict the putative recombination events leading to the His⁺ Pro⁺ recombinants in the M146 x ZX7 cross (c), the His⁺ Pro⁺ recombinants with same telomeres (d) and with mixed ends (e) in the 1190 x ZX7 cross, and the Ura⁺ Pro⁺ recombinants in the same cross (f). The chromosomes of S. coelicolor (SC) and S. lividans (SL) are aligned with respect to their corresponding AseI fragments and genetic markers (h - hisA1; p - pro-2; d - agarase, u - uraA1). The homology between the AseI linking clones of the two chromosomes (Leblond, et al., 1993) is indicated by the connecting grey lines. Those AseI fragments whose origins could be putatively assigned size-wise without ambiguity are boxed: filled box for presence and open boxes for absence in the recombinant chromosomes. In panel (c), the presence of the C fragment of S. coelicolor was clear in recombinant No. 5, but not in the other three. In panel (e), the presence of intact A fragment of S. lividans could not be established for lack of sufficient resolution [panel (b)], but the presence of its end was identified by hybridization (not shown) and was so indicated. The genetic markers selected in the crosses are indicated by the filled triangles, and the unselected markers predominant in the recombinants are indicated by the open triangles. Based on the restriction fragments and markers identified, the structures of the recombinant chromosomes are traced along the parental chromosomes with the putative crossovers between the parental chromosomes indicated by the vertical lines.

Fig. 4. Models for obligated double crossovers during recombination of Streptomyces chromosomes. (a) Recombination between an intact linear chromosome (left) and a partial chromosome with a telomere (right). Both odd number (shown) and even number of crossover are legitimate, generating recombinant chromosomes with mixed ends or same ends, respectively. (b) Recombination between an intact linear chromosome (left) and an internal chromosomal sequence (right). Even-numbered crossovers are necessary to give rise to a complete recombinant chromosome. (c) Recombinations between two linear chromosomes with free ends. Both odd and even numbers of crossovers are allowed, giving rise to mixed ends and same ends, respectively. (d) Recombination between two intact linear chromosomes with strongly interacting telomeres. Even numbers of crossovers are necessary for separation of the recombinant chromosomes. Two pairs of terminally located alleles, a⁺/a^- and z⁺/z^-, are indicated to illustrate linkage analysis. The TPs are represented by the filled circles.

REFERENCES