Microbiology (1999),
145, 2183-2202.
© 1999 Society for General Microbiology
|
Molecular Genetics of Streptomycetes |
Forty years of genetics with Streptomyces: from in vivo through in vitro to in silico
David A. Hopwood1
John Innes Centre, Norwich Research Park, Colney, Norwich NR4 7UH, UK1
Tel: +44 1603 452571. Fax: +44 1603 456844. e-mail: hopwood{at}bbsrc.ac.uk
Keywords: Streptomyces, genetics, antibiotics, chromosome-linear
 |
Beginnings: the A3(2) strain
|
|---|
As an undergraduate at Cambridge in the early 1950s, I had found genetics the most exciting part of my courses. This was the main motivation for choosing Botany as my Part II (final year) option in the academic year 1953/54. Granted, there was the Department of Genetics at Cambridge, headed by none other than Ronald Fisher, the father of biological statistics and a true genius, but an advanced qualification in mathematics was a de facto entry ticket to Part II Genetics, reflecting the strong mathematical basis of research in that Department. In the Botany School, ‘biochemical genetics’ was being studied – using Neurospora crassa – by Harold Whitehouse and Lewis Frost. During the Part II course in Botany, they far-sightedly emphasized the power and promise of microbial genetics for the future understanding of gene structure and function. In choosing a PhD project to start in October 1954 I heeded their advice. One of the topics on offer was Streptomyces. Here was a group of microbes apparently ‘intermediate between bacteria and fungi’, and so they could not fail to be interesting from a genetic point of view, given the clear differences that were emerging between the genetics of the few bacteria so far studied and the genetics of fungi and higher organisms. In the Pneumococcus, Escherichia coli and Salmonella typhimurium, incomplete genomes were transferred from donor to recipient strains, by any of three bizarre processes (transformation, conjugation and transduction), to yield incomplete zygotes, whereas in fungi and higher organisms (except those that were asexual) life cycles, including a complete diploid stage and meiosis, were the rule.
To get me started, Harold Whitehouse lent me a copy of Waksman’s book on the actinomycetes (Waksman, 1950
), newly purchased from Heffers’ bookshop, and Lewis Frost handed over half a dozen cultures of Streptomyces that he had obtained from his friend George Floodgate at the then Royal College of Technology in Glasgow (now Strathclyde University). I streaked them out: several grew well, and one produced a striking blue pigment. I could therefore (I thought!) name it Streptomyces coelicolor. As a bonus, pigmentation might in due course make a valuable genetic marker, as it had in Aspergillus nidulans. I chose this culture (strain 204F) and set about isolating auxotrophic mutants in order to look for genetic recombination à la Lederberg & Tatum (1946 ). Pretty soon, I could grow pairs of mutants together and select rare prototrophs from the progeny spores. In crosses of two doubly marked strains, non-selected markers from each parent segregated amongst the progeny, indicating a genuine process of gene reassortment rather than simply some kind of heterozygote formation. However, by early 1955 problems had arisen: the cultures started to grow erratically and became ‘friable’ (Fig. 1). I diagnosed phage infection, but could not cure it. I therefore wrote to Dagny Erikson, at the University of Aberdeen, to ask for cultures of S. coelicolor. She had published extensively on variation in cultures that she had derived from Stanier’s agar-liquefying strain (Stanier, 1942 ), including some derived by using a micromanipulator to isolate individual spores from the same chain (Erikson, 1948 ). Her initial response was disappointing: she had retired to Kincardineshire to raise a family (as the new Mrs A. E. Oxford) and doubted whether any viable cultures remained. However, she would see if Dr D. M. Webley could help. A week later, a further letter arrived: ‘Dr Webley informs me that his culture of Streptomyces coelicolor is not in good condition. I have found an old sterile soil ampoule which was sown with the blue-pigmented, agar-liquefying strain A3(2). I cannot guarantee it, but hope it may still be viable and pure.’ Indeed it was! I shook some soil particles from the ampoule onto an agar plate and beautiful blue colonies grew out. I made auxotrophic mutants and they were stable, and grew perfectly. They produced prototrophic recombinants (Fig. 2) when grown in pairs (Hopwood, 1957 ). Thus was the A3(2) strain launched on its genetic career. Meanwhile, the true phylogenetic relationships of Streptomyces needed looking into.
View larger version (88K):
[in this window]
[in a new window]
|
Fig. 1. Colonies of S. coelicolor strain 204F: left, normal sporulating colonies of the original strain; right, a ‘friable’, asporogenous culture, possibly caused by phage infection. Reproduced from Hopwood (1958 ).
|
|
View larger version (92K):
[in this window]
[in a new window]
|
Fig. 2. Some of the first recombinants selected from a four-factor cross between marked derivatives of S. coelicolor A3(2), metA2xhisA1 argA1 strA1. Left, recombinants on a medium containing histidine (selection for met+/arg+); right, recombinants on a medium containing arginine, methionine and streptomycin (selecting his+/str). Note the slight background of parental growth on the left-hand plate caused by cross-feeding of the parental strains and its absence on the right-hand plate because one of the parents is killed by the streptomycin. Reproduced from Hopwood (1958 ).
|
|
|
Streptomycetes are bacteria
|
|---|
Early in the development of microbiology, there were several descriptions of novel organisms that are now included in the actinomycetes, to which Streptomyces belongs. Hansen (1874 ) saw the causal agent of leprosy in tissue samples from human patients and called it Bacillus leprae; Cohn (1875 ) named as Streptothrix foersteri a filamentous organism from human tear ducts; and Harz (1877 ) described Actinomyces bovis in the lesions of ‘lumpy jaw’ in cattle. It was probably the naming of this last organism as ‘ray fungus’, because of the tendency of the elongated, branching cells to grow radially from the centre of the lesion, that later contributed to the confusion surrounding the taxonomic position of the Actinomycetales; the renaming of Hansen’s bacillus as Mycobacterium (‘fungus rodlet’) and Cohn’s Streptothrix (‘twisted hair’) as Streptomyces (‘twisted- or chain-like fungus’) did not help.
By the early 1950s, few microbiologists regarded the actinomycetes as fungi, but many still thought of them as intermediate between fungi and bacteria. For Streptomyces, there were biochemical pointers to a bacterial affinity – the cell-wall composition resembled that of typical Gram-positive bacteria, and they were sensitive to specifically anti-bacterial antibiotics – but their cellular architecture was unclear. In Cambridge, through a mutual friend (Martin Canney), I was lucky enough to meet the brilliant electron microscopist Audrey Glauert who, with Ernst Brieger, was studying the fine structure of Mycobacterium strains at the Strangeways Laboratory. This was the start of a long and rewarding collaboration during which we showed, among other things, that S. coelicolor lacks a nuclear membrane (Fig. 3) and so is, by definition, a bacterium (or ‘prokaryote’ to use a later term). Nevertheless, it was clearly very different morphologically from the simple, rod-shaped bacteria that had hitherto been studied genetically, so the idea that it might reveal novel genetic phenomena was still very much alive.
View larger version (63K):
[in this window]
[in a new window]
|
Fig. 3. A thin section of a germinated spore (S) of S. coelicolor A3(2) with its germ tube, together with two other spores, showing prokaryotic cellular architecture. The hypha is bounded by the cell wall (W), under which lies the plasma membrane (PM), which is continuous with peripheral pockets of membranous material (P). The fibrillar nuclear material (N) extends from the spore into the germ tube, and from the hypha into a side branch and is not delimited from the cytoplasm by a membrane. Intracytoplasmic membranous regions (CM) – later called mesosomes – are also seen. There is an irregular layer of material (M) on the outer surface of the wall. See also Glauert & Hopwood (1961 ).
|
|
|
Does Streptomyces have a life cycle?
|
|---|
The very first issue of the Journal of General Microbiology (forerunner of Microbiology) contained two papers by Dagny Erikson on ‘Differentiation of the vegetative and sporogenous phases of the actinomycetes’ Erikson (1947a ,b ), and also one by Emmy Klieneberger-Nobel on ‘The life cycle of sporing Actinomyces as revealed by a study of their structure and septation’ (Klieneberger-Nobel, 1947 ). In the latter was described, for four strains of what we would now call Streptomyces, a life cycle consisting of two growth phases: a ‘primary’ or ‘substratum’ mycelium gave rise, by a process of cellular fusion (and implied fusion of ‘chromosomes’) in special hyphal aggregates, to ‘initial cells’; these then produced the sporulating, ‘secondary’ or ‘aerial’ mycelium by a germination-like process (Fig. 4).
If true, such a life cycle, with an implied regular alternation of haploid and diploid phases, would obviously have had profound implications for Streptomyces genetics. However, it could not be substantiated. Erikson (1955 ) had found that aerial hyphae of the A3(2) strain arose as simple side-branches of vegetative hyphae anywhere over the colonies. I confirmed this (Fig. 5), and criticised the interpretation by Klieneberger-Nobel of her own observations, concluding that ‘there appear to be no objective reasons for Klieneberger-Nobel’s interpretation of those of her photographs which are claimed to show initial cells’ (Hopwood, 1960 ). In hindsight, this was perhaps a rather impertinently expressed criticism of a senior microbiologist by a brash neophyte, but the point appeared to be in need of emphasis, especially because the idea of initial cells as a stage in the life cycle was beginning to be assumed in the literature (e.g. McGregor, 1954 ; Dickenson & MacDonald, 1955 ).
View larger version (44K):
[in this window]
[in a new window]
|
Fig. 5. Phase-contrast photomicrograph showing a spore (left) of S. coelicolor A3(2) which has germinated to produce a vegetative mycelium on which two aerial hyphae (growing upwards in the photograph), one already going on to sporulate, have arisen as simple side branches. Reproduced from Hopwood (1960 ).
|
|
|
Genetic studies in other laboratories
|
|---|
As is often the case in science, I was not alone in my work: the time was evidently ripe to look for genetic recombination in Streptomyces. It transpired that at least five other groups had started similar work at about the same time, all unbeknown to each other (Sermonti & Spada-Sermonti, 1955 ; Alikhanian & Mindlin, 1957 ; Bradley & Lederberg, 1956 ; Braendle & Szybalski, 1957 ; Saito, 1957 ). Remarkably, two of them had also independently chosen blue-pigmented strains of S. coelicolor, probably for the same reason as I had done. While the possible intermediate phylogenetic position of Streptomyces had been a motivation for several groups, as it had been for me, most also had a desire to develop rational breeding methods in the group of microbes that were beginning to establish themselves as the world’s most prolific antibiotic producers (Szybalski, 1959 ); this had not been a factor in my own case because I knew next to nothing about antibiotics. I was forced to take an interest in them only later when the genetics of the blue pigment started to open up. Most of the ‘competition’ dropped out after about 1960, while I developed a fruitful collaboration with the group who had published first, Giuseppe Sermonti and his wife Isabella Spada-Sermonti in Rome. They soon adopted the A3(2) strain, with its linkage map and my growing collection of multiply marked mutants and recombinants (see later section). Eventually, our paths diverged and the Sermontis moved to other fields in the 1970s.
|
Taxonomy of the A3(2) culture
|
|---|
I have to ‘blame’ Sermonti for the opprobrium that we later attracted for persisting in referring to the A3(2) strain as ‘S. coelicolor’. In 1959, Kutzner & Waksman (1959 ) concluded that Streptomyces strains that make blue pigments represent two distinct taxonomic groups (even the pigments are chemically quite different), and that a strain described by Müller had priority for the name S. coelicolor; strains resembling A3(2) should be called ‘S. violaceoruber’. I was convinced by the argument, and Audrey Glauert and I started to use the new name in our cytological papers (e.g. Glauert & Hopwood, 1961 ; Hopwood & Glauert, 1961 ). However, Sermonti persuaded me that there was no guarantee the name would not be altered again and that we would confuse geneticists by changing horses in mid-stream.
|
Establishing a linkage map
|
|---|
Having concluded that Streptomyces are bacteria and, like other bacteria, show no objective evidence of a life cycle involving anything other than asexual reproduction, the task was to try to establish a system that could illuminate the genetics of S. coelicolor A3(2) purely by analysis of recombinants produced in pairwise mixtures of auxotrophic strains. In trying to do so, I was mindful of the problems that bedevil any deduction of the linkage relationships of bacterial genes when recombinants have to be selected from a huge excess of asexually produced progeny of the two parental groups. The consequence of this is that the frequencies of particular classes of recombinants in respect of the non-selected markers do not simply reflect the distance apart of these markers, which one is trying to establish, but also the positions of these genes with respect to the selected and counter-selected markers, which might lie between the non-selected markers, or outside of them, or one between and one outside. These problems had been dramatically illustrated in the paper by Lederberg et al. (1951 ), in which the linkage map of E. coli K-12 (though presumably not the chromosome itself) had to be represented as a branched structure, before an understanding of the donor and recipient roles of the parents in the cross allowed sense to be made of the linkage data.
My solution to the problem came to me out of the blue one afternoon when I was explaining my project to two German colleagues from St John’s College over tea in the Botany School, one a linguist and the other an economist (thereby doing nothing to dispel the idea prevalent among non-scientists that science advances by the Eureka principle!). This was the four-on-four cross. In it two parents each have two markers, which can be either selected or non-selected. Recombinants are recovered from equal numbers of total progeny on four media, each making one of the four possible selections of one marker from each parent and leaving the second marker of that parent non-selected. Because there are two non-selected markers, four recombinant genotypes can grow on each medium; their frequencies are determined by picking and classifying samples of the colonies. Some of the genotypes grow on more than one of the media, while others are unique to a single medium. Taking them all together, they include one or both members of each of the seven possible complementary pairs of recombinant genotypes, leaving only the parental classes (which are, of course, the same as the vast excess of asexual progeny) unmeasured. In this way relative linkage distances could be measured and used to construct a map.
The first version of the map (Hopwood, 1958 ) had just six loci, in two linkage groups, and this was soon extended to 15 loci, with the (temporary) appearance of a third linkage group represented by a single gene (Hopwood, 1959 ). With the discovery of heteroclones – colonies deriving from partially diploid plating units, each of which gave rise to a population of haploid recombinants that could be analysed non-selectively (Sermonti et al., 1960 ; Hopwood et al., 1963 ) – the map grew to 39 loci, still in two linkage groups (Hopwood, 1965a ). Soon, the map was firmly enough established for a simpler system of analysis to be employed. Selection was made for recombination between just two known points on the chromosome with other markers unselected. In appropriate crosses, the pattern of marker selection in one linkage group influenced, in very specific ways, the segregation of markers in the other linkage group; the interpretation was a single, circular linkage group that incorporated the two previously separate groups of genes (Hopwood, 1965b ).
The reason it had not been possible earlier to bridge the gaps between the two original linkage groups was that they were separated by long, very sparsely marked segments. These so-called ‘silent regions’ were devoid even of a general class of temperature-sensitive mutants, as described in a paper (Hopwood, 1966a ) that acknowledged for the first time the input of Helen Ferguson (later Wright, still later Kieser), who has contributed so much to our 35 year collaboration. These regions remained for long a striking feature of the S. coelicolor genome and would be interpreted only some 30 years later (see later section) when Helen was able to map the chromosome physically (Kieser et al., 1992 ).
As Frank Stahl (1967 ) had taught us, this map circularity did not necessarily mean that the chromosome itself was a circle: merozygosity, leading to the requirement for even-numbered crossovers to generate recombinants between a complete chromosome from one parent and a fragment from the other (except in the rare case when the fragment included a chromosome end), could have been a sufficient explanation for circularity of the genetic map even if the chromosome were linear (Fig. 6). For technical reasons, it was not possible to obtain the kind of direct physical evidence that had established chromosome circularity in E. coli (Cairns, 1963 ), so I tried at least to distinguish between linearity on the one hand and circularity or circular permutation on the other by analysing several hundred heteroclones selected from the same cross. It turned out that the region of heterozygosity in a particular heteroclone could cover any continuous arc of the linkage map. On the assumption that this region represented an uninterrupted segment of the chromosome, the conclusion was that the chromosome lacked constant ends (Hopwood, 1966b ). It was provisionally assumed to be circular pending direct evidence, but might have been circularly permuted instead. It was a very long time before we knew more than this!
View larger version (3K):
[in this window]
[in a new window]
|
Fig. 6. Diagram to illustrate how, in a merozygote (partial diploid) with a complete (linear) chromosome from one parent and an incomplete chromosome from the other (three examples are shown, with the incomplete chromosome coming from one parent or the other and representing various regions of the chromosome), only even-numbered crossovers (double crossovers are shown here) yield viable recombinants with a complete chromosome (except in the rare cases where the incomplete chromosome includes an end). It follows that almost all recombinant progeny inherit both chromosome ends from the same parent; in other words the ends are 100% linked, thereby yielding map circularity.
|
|
|
The three phases of Streptomyces genetics
|
|---|
In vivo genetics was the order of the day for all organisms until the recombinant DNA revolution, starting in 1974, opened the way for the in vitro era, and changed genetics for ever. Methods for routine DNA sequencing, developed in 1977, led in turn to the in silico phase, based primarily on the explosion of databases and ever more sophisticated sequence comparisons. It took a little longer for these developments to take hold in Streptomyces. Although the phases overlapped of course, I shall regard the in vivo phase as occupying the rest of the 1960s and the 1970s; the in vitro phase as dominating the 1980s; and the 1990s seeing the rise of in silico Streptomyces genetics.
In skimming over developments in Streptomyces genetics since 1965, I have inevitably made subjective choices, not to say parochial ones [there is certainly a bias in favour of S. coelicolor A3(2)], even though I have tried to include in Tables 1–3 appropriate discoveries from other laboratories. I hope the choice will be viewed indulgently by those whose work I have omitted to mention.
The in vivo years
The major discoveries of the in vivo years are summarized in Table 1
. An intriguing feature of the chromosome first observed in the mid-1960s, but still awaiting evaluation, is the indication of an ancestral duplication of the genome. I noticed a tendency for genes for different steps in the same metabolic pathway to lie opposite one another on the circular linkage map. I proposed crossing-over between sister chromosomes in the same hyphal compartment to generate an ancestral double-sized circle, with all gene clusters duplicated on opposite sides of the map. Progressive loss of (most) duplicate genes would then leave some members of a cluster on each side (Fig. 7; Hopwood, 1967a ). It will be fascinating to see if the current genome sequencing project (see below) reveals the kind of ‘mirror-image’ patches of residual sequence identity that might be expected on such a model.
View larger version (17K):
[in this window]
[in a new window]
|
Fig. 7. A possible explanation for map symmetry in S. coelicolor A3(2). An ancestral (circular) chromosome (i) has three clusters of biosynthetic genes – say A for arginine, H for histidine, T for tryptophan – each consisting of three genes, a, b, c. A crossover between sister chromosomes in the same mycelial compartment gives rise to a double-sized circle, on which duplicate clusters lie opposite to each other (ii). This dimer would be unstable: a further crossover would restore it to the monomeric state. However, the dimer could become stabilized by the accumulation of even a small number of recessive lethal mutations which would render the monomeric chromosomes inviable. Further mutations, including deletions, would then arise to leave, say, one copy of the A cluster with genes a and c, and the opposite copy with gene b (iii). Sometimes the whole of one cluster (for example T) might be lost, leaving an ‘unpaired’ cluster. See also Hopwood (1967a ).
|
|
1967 also saw the publication of a discovery that few people (including me) appraised correctly at the time – partly because it was made in the Soviet Union and reported in the Russian literature – but which is now seen as seminal. This was the discovery of A-factor as a diffusible molecule required, in minute concentrations, to activate streptomycin production and sporulation in Streptomyces griseus (Khokhlov et al., 1967 ). A tour de force of natural product chemistry by the Khokhlov group led to chemical characterization of this hormone-like molecule as a novel 
-butyrolactone (Fig. 8; Kleiner et al., 1976 ). The later rediscovery of the Russian work, and its enormous development in S. griseus, Streptomyces virginiae and S. coelicolor, primarily by the groups of Teruhiko Beppu and Sueharu Horinouchi and of Yasuhiro Yamada (Hara & Beppu, 1982 ; Yamada et al., 1987 ), has been one of the most significant facets of Streptomyces genetics over recent years (e.g. Onaka et al., 1998 ).
A second highly significant development also took place in Moscow in the late 1960s. This was the penetrating work of Natalia Lomovskaya and her collaborators on the 
C31 temperate bacteriophage (Fig. 9; Lomovskaya et al., 1970 ), which also launched its favoured host, Streptomyces lividans 66, on the world stage. Her generous sharing of her work and strains with Western scientists – including in particular Keith Chater at John Innes – at a difficult time for Russian geneticists has had far-reaching benefits.
View larger version (100K):
[in this window]
[in a new window]
|
Fig. 9. The C31 temperate actinophage (kindly provided by T. Clayton, The John Innes Centre, Norwich, UK).
|
|
Meanwhile, the varying fertility status of substrains of A3(2) was beginning to be interpreted. It became apparent that, just as in E. coli, plasmid sex factors were at work and that they had assumed different states in different strains: autonomous, integrated, primed and absent. I had noticed that when progeny selected from my early crosses between mutants of the wild-type were crossed together they produced higher frequencies of recombinants than these early crosses themselves, but had not realized the significance of this difference. Sermonti had done so, and had begun to study the phenomenon (Sermonti & Casciano, 1963 ). In Glasgow, we obtained a particularly surprising result in crosses between so-called NF and UF strains (later interpreted as having an integrated SCP1 plasmid and no SCP1 respectively: Vivian & Hopwood, 1970 ; Vivian, 1971 ). These crosses produced recombinants so abundantly that they could be analysed non-selectively, a rare occurrence in a bacterium (Hopwood et al., 1969 ).
The plasmid nature of SCP1 was deduced genetically. In spite of some raised eyebrows, because we had not identified the element physically, this did not render the conclusion invalid in view of the original definition of the word ‘plasmid’, which did not refer to its physical nature (Lederberg, 1952 ). However, we needed to isolate SCP1, and it long resisted isolation, notwithstanding valiant attempts by Janet Westpheling and others (see Hopwood et al., 1979 ). The reason, discovered nearly ten years later by Haruyasu Kinashi (Kinashi et al., 1987 ), who exploited the newly invented pulsed-field gel technology, is that SCP1 is both large (350 kb) and linear. With hindsight, we should have taken more seriously the possibility that SCP1 was linear, because linear plasmids small enough to be revealed by conventional technology had been found as early as 1979 (Hayakawa et al., 1979 ). This was another momentous development, the full significance of which became apparent only much later when the widespread occurrence in Streptomyces of linear replicons, including the chromosome, was realized.
Alan Vivian (1971 ) had found that SCP1+ strains inhibited SCP1- strains through the agency of a diffusible molecule, but only later was this found to be a small molecule (i.e. not a protein like a bacteriocin) with a wide spectrum of activity (Fig. 10) and so it could be called an antibiotic (Kirby et al., 1975 ), later identified as methylenomycin (Wright & Hopwood, 1976a ). This, the first example of genetic localization of antibiotic-biosynthetic genes, led to much speculation that plasmid determination of antibiotic production might be widespread, if not typical (e.g. Umezawa, 1977 ). Ironically, methylenomycin remains, after 20 years, the only unambiguous example of a Streptomyces antibiotic specified by a plasmid. The very next example – actinorhodin in S. coelicolor (Wright & Hopwood, 1976b ) – implicated chromosomal genes, later shown to be clustered (Rudd & Hopwood, 1979 ), and this instead has turned out to be the paradigm.
View larger version (95K):
[in this window]
[in a new window]
|
Fig. 10. Assay of methylenomycin production. Agar plugs of five SCP1+ and two SCP1- S. coelicolor A3(2) strains were placed on a Petri plate spread with a lawn of Bacillus mycoides. After incubation, the SCP1+ plugs are surrounded by zones of inhibition of the lawn.
|
|
In parallel with the mapping of antibiotic-biosynthetic genes, developmental mutants, defining the other special feature of Streptomyces biology, were also mapped in the early 1970s. One of them – S48, later named bldA (see below) – had actually been used as a textbook example of genetic mapping even earlier (Hopwood, 1967b ), but systematic classification and mapping of whi (white colony) mutants came a little later (Hopwood et al., 1970 ; Chater, 1972 ), and the bld mutants were given similar treatment in 1976 (Merrick, 1976 ).
Meanwhile, a second plasmid had surfaced in S. coelicolor; this proved to be a conventional, moderate-sized, CCC molecule, amenable to isolation by a fairly standard protocol (Schrempf et al., 1975 ). SCP2 was largely responsible for the fertility in crosses that could not be attributed to SCP1 (Bibb et al., 1977 ); a variant form of the plasmid, called SCP2*, was an even better sex factor. The discovery of SCP2* was an essential stepping stone towards introducing plasmid DNA into Streptomyces hosts by transformation, thereby opening the way for gene cloning in these organisms. The transformation procedure devised (Bibb et al., 1978 ) depended on the earlier development of protoplast fusion, which gave rise to enormously high frequencies of genetic recombination without the need for a plasmid vector (Hopwood et al., 1977 ; Baltz, 1978 ). These studies, in turn, could not have been made without the pioneering work of Masanori Okanishi and his colleagues, who painstakingly investigated the factors necessary for good protoplast formation and, even more crucially, their regeneration (Okanishi et al., 1974 ).
Plasmid biology was also beginning to develop as the 1970s were coming to an end. The marker used by Bibb et al. (1978 ) to track transformants by SCP2* was ‘pock’ formation: transformants receiving SCP2* gave rise to small, circular areas of retarded growth in a lawn of the plasmid-free strain. Originally this effect was assumed to be an actual killing of the strain by the mating process, and was likened to lethal zygosis in E. coli (Skurray & Reeves, 1973 ). An even earlier notion, when pocks were seen in matings between S. coelicolor carrying SCP1 and S. lividans (Hopwood & Wright, 1973a ), was that inhibition of the S. lividans lawn was due to the diffusible agent that was subsequently identified as methylenomycin. Only later, when plasmid mutants altered in pock formation were isolated (Fig. 11: Kieser et al., 1982 ), was the phenomenon interpreted (we hope correctly, but direct proof is still lacking) as a manifestation of the migration of a conjugative plasmid within the mycelium of the recipient. This process was described as plasmid ‘spreading’ and became capable of molecular analysis only in the in vitro years.
View larger version (135K):
[in this window]
[in a new window]
|
Fig. 11. Pocks produced by the growth of a few plasmid-containing spores of a strain of S. lividans in a background of a plasmid-free S. lividans culture. Left: pocks produced by a pIJ101 derivative with wild-type ‘spread’ genes; each pock is a few millimetres in diameter and represents the area of the indicator culture into which the plasmid has migrated. Right: pocks produced by a plasmid mutated in the ‘spread’ genes for putative intramycelial migration; the much reduced size of the pocks was the first clue to the existence of such genes. See also Kieser et al. (1982 ).
|
|
Pocks meanwhile led to the discovery of a further major class of Streptomyces plasmids. The first example, SLP1, was revealed when pocks were found in a lawn of S. lividans after contact with an S. coelicolor strain devoid of SCP1 and SCP2 (Bibb et al., 1981 ). The pocks yielded a family of SLP1 plasmids of varying size. SLP1 was evidently integrated in the S. coelicolor chromosome and could loop out to form a conjugative plasmid. A large family of such elements was soon discovered, of which the most extensively studied, along with SLP1 (e.g. Shiffman & Cohen, 1993 ), is pSAM2 from Streptomyces ambofaciens (e.g. Sezonov et al., 1995 ).
As the 1970s came to an end, generalized transduction was added to the other two modes of gene transfer – by plasmid-mediated conjugation and by transformation (albeit artificial) – when Colin Stuttard described a phage, SV1, that could transduce markers in Streptomyces venezuelae (Stuttard, 1979 ). Unfortunately, SV1 did not work in any other species; and the second example of general transduction, by SF1 in Streptomyces fradiae (Chung, 1982 ), was also taxonomically circumscribed. Only very recently has a system of general transduction for S. coelicolor and S. lividans been developed (J. Westpheling, personal communication), thereby promising to fill a significant gap in the genetic tool-kit.
The in vitro years
The major achievements of the in vitro years are summarized in Table 2. The 1980s opened momentously for Streptomyces genetics with the first reports of gene cloning in S. coelicolor and S. lividans (Bibb et al., 1980 ; Suarez & Chater, 1980 ; Thompson et al., 1980 ). Of the genes cloned in these first experiments, several conferring antibiotic resistances soon established themselves as versatile tools in vector development; the tsr gene, for thiostrepton resistance (Thompson et al., 1980 ), is perhaps the most famous, having been used as a selective marker on a multitude of Streptomyces cloning vectors (a mixed blessing perhaps, because this antibiotic is not the easiest to come by!).
The developments in Streptomyces in vitro genetics that followed during the 1980s may be divided into two broad groups. One includes the beginnings of an understanding of basic features of the molecular biology and physiology of the organisms, such as gene and operon structure, and the control of transcription, translation and primary metabolism; the other represents the use of gene cloning to understand the molecular genetics of some of the special features of Streptomyces biology, especially antibiotic production and the developmental cycle.
In the first category, developments were coming thick and fast by the mid-1980s. Streptomyces promoter sequences were isolated (Bibb & Cohen, 1982 ) and soon after that, the first Streptomyces gene was sequenced (Thompson & Gray, 1983 ). Not surprisingly, this gene sequence revealed an extraordinarily high proportion of codons ending in G or C, and this bias, typical of Streptomyces genes, was exploited in the frame program (Bibb et al., 1984 ), which remains an extremely useful tool for finding genes (and frameshift errors!) in DNA sequences. Soon, promoters were sequenced and found to be both heterogeneous and abundant: many genes turned out to have more than one transcription-start site (e.g. Bibb et al., 1985 ; Buttner et al., 1987 ), a finding that is still not understood in any detail but which may reflect the need for subtly modulated expression at different stages in the life cycle or in different physiological states.
Pioneering studies revealed multiple forms of RNA polymerase holoenzyme in S. coelicolor (Westpheling et al., 1985 ; Buttner et al., 1988 ), establishing this as a potential strategy for differential regulation of different subsets of genes in Streptomyces, following the Bacillus subtilis paradigm (Losick & Pero, 1981 ). The number of different sigma factors in S. coelicolor has increased steadily over the years, including the remarkable and still only partially understood presence of four sigma factors that strongly resemble the vegetative E. coli and B. subtilis form (Tanaka et al., 1988 ); only one of these is essential (Buttner, 1989 ). Currently, the number of known or inferred alternative sigma factors in S. coelicolor runs to at least 20 (M. J. Buttner, G. Kelemen & M. Paget, personal communication): even more than in B. subtilis.
Another surprising discovery was the relative abundance of Streptomyces genes that must be translated without a conventional Shine–Dalgarno sequence, because there is no untranslated leader on the mRNA before the start codon (Janssen et al., 1989 ). This provided one of the rather rare examples of a basic aspect of molecular biology that was brought to prominence in Streptomyces rather than in E. coli.
The first examples of Streptomyces operons to be rigorously defined reflect aspects of primary metabolism, and specify the proteins needed for the utilization of galactose and glycerol (Fornwald et al., 1987 ; Smith & Chater, 1988 ). Both are subject to specific induction as well as generalized carbon-catabolite repression by glucose (Mattern et al., 1993 ; Hindle & Smith, 1994 ). This global regulatory control has itself turned out to differ fundamentally in mechanism from the E. coli paradigm, with an involvement of glucose kinase, but not of cyclic AMP (Hodgson, 1982 ; Angell et al., 1994 ).
In the second category, the 1980s saw the beginning of our understanding of the molecular genetics of antibiotic production and its manipulation. The first antibiotic-biosynthetic genes were cloned, by a variety of methods (Chater & Bruton, 1983 ; Feitelson & Hopwood, 1983 ; Gil & Hopwood, 1983 ), and soon genes for a whole pathway, the act genes for actinorhodin biosynthesis, were isolated and expressed in a different Streptomyces host (Malpartida & Hopwood, 1984 ). Just as loss of the blue colour of actinorhodin had been instrumental in the isolation and in vivo mapping of antibiotic-pathway genes (Rudd & Hopwood, 1979 ), so its reacquisition was crucial to the isolation and in vitro mapping of the genes (Malpartida & Hopwood, 1984 , 1986 ). Soon, colour was exploited again, this time to produce a hybrid antibiotic by genetic engineering; when specific segments of the act gene cluster were introduced on a plasmid vector into the producer of the brown antibiotic medermycin, a beautiful purple culture arose and was found to be making mederrhodin, a compound with structural features of both medermycin and actinorhodin (Hopwood et al., 1985a ).
While cloning the act genes demonstrated close linkage of biosynthetic structural genes, the finding that Streptomyces parvulus containing the act cluster produced actinorhodin without killing itself (Malpartida & Hopwood, 1984 ) provided indirect evidence that a resistance gene had also been cloned; the basis of actinorhodin self-resistance, while probably in some way reflecting antibiotic export (Bystrykh et al., 1996 ), is still not well understood. In contrast, the methylenomycin gene cluster, on SCP1, was proved to contain resistance as well as biosynthetic and regulatory genes. This followed from the recognition of a specific methylenomycin-resistance gene – the very first Streptomyces gene to be cloned in Streptomyces (Bibb et al., 1980 ) – as part of the DNA that caused methylenomycin production in a non-producing host, and overproduction when a deduced regulatory region was inactivated (Fig. 12; Chater & Bruton, 1985 ).
On the chromosome front, the 1980s saw the flowering of a whole new subfield of Streptomyces genetics stemming from the report by Robinson et al. (1981 ) of the remarkable amplification of specific DNA segments. This phenomenon, which was later found to be associated with large (up to 1000 kb) deletions, has been studied in many strains by several groups (e.g. Schrempf, 1982 ; Ono et al., 1982 ; Altenbucher & Cullum, 1984 ). The apparent association of deletion and amplification cycles with the end regions of the linear chromosome, although probably not in an obligate way, has recently given the topic a new impetus (e.g. Lin & Chen, 1997 ; Volff & Altenbuchner, 1998 ).
Two further techniques were added to the toolbox for Streptomyces genetic manipulation at the end of the 1980s. The first Streptomyces transposon was identified in 1987. This was a Tn3-type element found in S. fradiae when it jumped on to the plasmid prophage of SF1, the transducing phage already referred to (Chung, 1987 ). A second was engineered from a naturally occurring insertion sequence in S. lividans (Solenberg & Burgett, 1989 ). The use of transposons in Streptomyces has lagged behind that in many other bacteria. However, the most promising seem now to be those isolated from other bacteria, such as Mycobacterium (Smith & Dyson, 1995 ), or even Tn5 (Volff & Altenbuchner, 1997 ). The second major advance was the demonstration of conjugal transfer of plasmids from E. coli to Streptomyces (Mazodier et al., 1989 ). The trick was to provide E. coli with a bifunctional plasmid containing a copy of oriT, as well as with mobilizing genes on the chromosome or on a second plasmid. This system has become a very useful one for the straightforward transfer of genes into Streptomyces (e.g. Bierman et al., 1992 ).
As a manifestation of all the activity in Streptomyces in vitro genetics in the 1980s, as well as the continued use of in vivo techniques, the John Innes Manual was published in 1985 (Hopwood et al., 1985b ). Judging by the number of research papers citing its use over the next decade, it filled a major gap. Now it is badly out of date, but we hope that the second edition, in preparation since 1994, will appear before the next millennium dawns (T. Kieser, M. J. Bibb, M. J. Buttner, K. F. Chater & D. A. Hopwood, unpublished).
The in silico years
The main developments of the in silico years are summarized in Table 3. If we define in silico genetics as the drawing of conclusions about the functions of genes from comparisons between a newly determined DNA sequence and sequences already in the databases, this phase of Streptomyces genetics probably began with the sequencing of the S. coelicolor bldA gene in 1987. Mutations in bldA abolish aerial-mycelium formation (hence the name bald) and also prevent production of the four known S. coelicolor antibiotics under certain nutritional conditions; this phenotype was so pleiotropic that the gene was thought likely to act at a central point in decision-making during the life cycle. Sure enough, the sequence showed that bldA could not encode a protein but rather would produce a typical tRNA with the anti-codon UUA (
TTA in DNA) (Lawlor et al., 1987 ). So was born the concept of a developmental switch that is dependent on the absence of the rare TTA codon for leucine from all genes required for vegetative growth, but present in a limited number of genes expressed post-exponentially (Leskiw et al., 1991 ). This idea was strikingly illustrated by the finding of a TTA codon in the 5' region of actII-ORF4, the gene for the pathway-specific activator of the actinorhodin cluster: mutation of this codon to TGA restored antibiotic production in a bldA background (Fernández-Moreno et al., 1991 ). The genome-sequencing project for S. coelicolor is providing a potentially complete inventory of genes containing TTA codons, and therefore making up a regulon controlled by bldA: the current prediction is about 100 genes (K. F. Chater, personal communication).
A second striking example of the power of in silico comparison came soon after: the sequence of the whiG gene, mutation of which causes the aerial mycelium to continue indeterminate growth and fail to switch to metamorphosis into chains of spores, showed it to encode a sigma factor (Chater et al., 1989 ), thereby establishing RNA polymerase heterogeneity as a strategy in Streptomyces development, recollecting the B. subtilis story (Losick & Pero, 1981 ). Again, sequencing and in silico analysis had short-circuited perhaps years of experimental attempts to elucidate the role of a developmental gene. This approach continues to bear fruit (e.g. Chater, 1998 ).
In the field of antibiotic biosynthesis, in silico genetics soon began to play an equally decisive role. Sequencing of the DNA regions responsible for the early carbon-chain-building steps towards the aromatic polyketides granaticin and tetracenomycin revealed that the enzymes involved – the first bacterial polyketide synthases to be studied by molecular genetics – were organized like those of a typical bacterial fatty acid synthase, strikingly confirming the idea that polyketide and fatty acid synthesis are homologous processes. The synthases were clearly so-called type II enzymes, consisting of separate subunits (Sherman et al., 1989 ; Bibb et al., 1989 ); this turned out to be the rule for aromatic polyketides (Fig. 13). In contrast, similar studies on the molecular genetics of erythromycin biosynthesis in Saccharopolyspora erythraea revealed a type I polyketide synthase organization, in which large proteins carry the different enzymic functions as a linear array of domains, an organization hitherto known only in eukaryotes. Even more remarkably, the sequencing revealed multiple sets, or modules, of domains (Fig. 14) that would together form an ‘assembly line’ for biosynthesis of the polyketide backbone (Cortes et al., 1990 ; Donadio et al., 1991 ). I have likened this discovery to that of the deduction of the mode of DNA replication directly from the Watson–Crick structure (Hopwood, 1997 ); the ‘assembly line’ model for complex polyketide biosynthesis had to be experimentally verified, but it arose directly from the determination of the gene sequences. These discoveries about the aromatic and complex polyketides in turn opened the way for the burgeoning field of ‘combinatorial biosynthesis’ of ‘unnatural natural products’ by the genetic engineering of polyketide synthases (McDaniel et al., 1993 , 1999 ; Khosla & Zawada, 1996 ; Hopwood, 1997 ).
View larger version (35K):
[in this window]
[in a new window]
|
Fig. 14. The modular erythromycin polyketide synthase and the ‘assembly-line’ model of biosynthesis leading to 6-deoxyerythronolide B, and on to erythromycin A through the action of tailoring enzymes. Reprinted with permission from Hopwood (1997 ). Copyright (1997) American Chemical Society.
|
|
Some other interesting examples of the power of in silico analysis in Streptomyces genetics are as follows: the first bacterial serine/threonine protein kinase (hitherto a eukaryotic preserve) was deduced from the sequence of afsK of S. coelicolor and S. lividans (Matsumoto et al., 1994 ); a new subfamily of sigma factors (Fig. 15), the ECF subfamily (for Extra Cytoplasmic Function), was discovered when the newly determined sequence of the S. coelicolor sigE gene was compared with other sequences in the databases (Lonetto et al., 1994 ); and the discovery of homology between the actII-ORF4 family of pathway-specific regulators (SARP, the Streptomyces Antibiotic Regulatory Proteins) and OmpR-type DNA-binding proteins uncovered a characteristic series of DNA motifs recognized by such proteins (Wietzorrek & Bibb, 1997 ).
View larger version (20K):
[in this window]
[in a new window]
|
Fig. 15. Phylogenetic relationships between the ECF subfamily and other RNA polymerase sigma factors. The new subfamily, with representatives in nearly all eubacteria, was recognized after the discovery of SigE from S. coelicolor A3(2) and greatly extended the span of the whole family of proteins. Reprinted with permission from Lonetto et al. (1994 ). Copyright (1994) National Academy of Sciences, USA.
|
|
The above examples of in silico Streptomyces genetics dealt with the sequencing of an individual gene or a group of clustered genes that had previously been identified by in vivo or in vitro genetics, rather than with the sequencing of an entire replicon followed by systematic in silico analysis of the complete sequence: in other words with ‘genomics’. The genomics phase of Streptomyces genetics began with the sequencing of some plasmids. Although a plasmid of 3706 bp, pSB24.1, had been sequenced as early as 1985 (Bolotin et al., 1985 ), the first sequence to make an impact was that of pIJ101 (Kendall & Cohen, 1988 ), because considerable in vivo and in vitro genetics had been done with this plasmid since its discovery in 1982 (Kieser et al., 1982 ). The 8830 bp sequence of pIJ101 revealed seven putative genes and immediately began to clarify the genetic control of plasmid transfer, both intermycelial and, putatively, intramycelial by the so-called spd (‘spread’) genes, as well as regulatory circuits for the control of replication and transfer (Fig. 16). The 41489 bp sequence of the model Streptomyces phage, C31, has been completed only very recently (Smith et al., 1999 ), and has led to many interesting conclusions. For example, all of the genes are orientated in the same direction except for one which, together with its promoter and operator, appears to have been inserted from elsewhere. There is much evidence for horizontal gene transfer involving other genes too, including perhaps genes shared by C31 and some eukaryotic viruses. Another fascinating find in C31 is a homologue of a plasmid ‘spread’ gene, suggesting a relationship between the putative intramycelial migration of plasmids and the movement of phage genomes between mycelial compartments during the lytic phase.
The Streptomyces chromosome, at 8 Mb, is enormous compared with most plasmids and phages so chromosomal genomics has taken longer to come on stream. A crucial enabling step was the use of PFGE to construct a restriction map of the S. coelicolor A3(2) chromosome for the two rare-cutting enzymes AseI and DraI, and to relate the physical to the genetic map (Kieser et al., 1992 ). The result was gratifying in that no examples were found in which gene order deduced from the results of in vivo crosses was contradicted by the physical evidence: even the relative spacing of the genes was quite well matched on the genetic and physical maps, except for an underestimation of the genetic length of the 3 o’clock ‘silent region’. This region was thus confirmed as a long stretch of DNA almost devoid of recognized genes, rather than a region of modest length in which the genetic distance was inflated by unusually frequent crossing-over. Physical linkage across both ‘silent regions’ (the shorter 9 o’clock one having in any case been partially filled by markers) appeared to have been confirmed, thereby solving the age-old question of chromosome circularity: the chromosome was concluded to be circular.
Hardly had the Kieser et al. (1992 ) paper been published than things took a dramatic turn. Carton Chen came on sabbatical to the John Innes Centre from the National Yang-Ming University in Taipei in the summer of 1992, with the first evidence for a linear chromosome in S. lividans, and he and Helen Kieser, with constant input from Yi-Shing Lin back in Taipei, clearly established this (Lin et al., 1993 ). Pretty soon, linearity was shown also for the S. coelicolor chromosome. How had this been missed? It turned out that a mistake had been made in positioning one of the DraI sites and so a point where apparent DraI and AseI sites coincided had been overlooked (this was the position of the chromosome ends); and long terminal repeats, later estimated to be >60 kb, at either end of the chromosome had led to hybridization of a probe taken from a point near to one end, to both ends, thereby mimicking a physical continuity across the 3 o’clock region. Carton Chen has vividly described the whole story, and especially the excitement of the hunt for the chromosome ends (Chen, 1997 ). The discovery has helped to change our ideas about bacterial chromosomes. It was also linked to elegant experiments on linear plasmids which proposed a model for Streptomyces chromosome replication in which conventional bidirectional replication takes place from a typical, centrally located oriC (Calcutt & Schmidt, 1992 ; Zakrzewska-Czerwinska & Schrempf, 1992 ), followed by ‘patching’ replication primed by proteins attached covalently to the free 5' ends of the chromosome to fill the gap left by removal of the RNA primer for the last Okazaki fragment o
TOP
Beginnings: the A3(2) strain
Streptomycetes are bacteria
