'

The EMBO Medal for 1991 has been awarded to Dr Patrick Stragier of the Institut de Biologie Physico-Chimique, Paris, France. Dr Stragier receives the medal for his work on the role of sigma factors in the control of sporulation in the soil bacterium Bacillus subtilis, about which he writes in the review beginning on the facing page.
The Medal is sponsored by the following companies: AKZO, Amersham, ASTRA, Becton-Dickinson, Boehringer Ingelheim, Boehringer Mannheim, Carlsberg Bryggerierne, Ciba-Geigy AG, F. Hoffman-La Roche AG, Sandoz AG, EMBL, Farmitalia, Glaxo, ICI, Kontron, LKB, Merieux, NOVO Industri A/S, Pharmacia, Sanofi, Sclavo, Senetek, Tuborgfondet.

The EMBO Journal vol.10 no.12 pp.3559-3566, 1991

EMBO MEDAL REVIEW

Dances with sigmas

Patrick Stragier
Institut de Biologie Physico-Chimique, 13 75005 Paris, France
rue

Pierre

et

Marie Curie,

Sporulation of Bacillus subtilis is developmental system

simple

When I started looking for my first lab just after finishing my theoretical studies at the University of Paris I was determined to work on the regulation of transcription. I had been enthralled by the elegance and subtlety of the mechanisms governing the control of lysogeny in bacteriophage lambda and I wanted to study a similar system. With this goal in mind I visited Luisa Hirschbein who had been working previously on RNA polymerase in Escherichia coli and had recently joined Pierre Schaeffer's laboratory at the Microbiology Institute of the University of Orsay. On that day I heard for the first time about an organism that looked even more fascinating than lambda, the sporulating soil bacterium Bacillus subtilis (Figure 1). When starved of nutrients (either carbon, nitrogen or phosphorous), B. subtilis undergoes a series of morphological changes which ultimately lead to the formation of a dormant spore (Figure 2) (reviewed in Losick et al., 1986). These spores can be seen under the light microscope as early as 'stage IV' (see Figure 2) and they accumulate a brown pigment which gives the colonies a characteristic appearance on plates. Because of their unpigmented phenotype many mutants had been selected which were blocked at specific stages during the sporulation process, and since generalized transducing phages were available, in addition to the remarkable property of B. subtilis of being spontaneously transformable, these mutations had been mapped on the chromosome and had defined several unlinked spo loci. Presumably these loci encoded products required for the elaboration of the spore and they had to be expressed in a tightly ordered fashion. Moreover, since the two compartments present during sporulation contained an identical chromosome issued from the last round of replication but behaved differently, the 'forespore' becoming the mature spore released by lysis of the 'mother cell' (see Figure 2), it could be speculated (although no evidence was then available) that different genes were expressed in the two compartments of the sporulating cell. Obviously, some quite complex mechanisms were involved in regulating transcription of the spo genes. On the day of my first visit I missed some of these exciting aspects of sporulation of B.subtilis, but I was immediately convinced that RNA polymerase played a central role in this developmental process and that it was the enzyme to study. Although I was captivated by Luisa's explanations I could not help noticing the intense activity of two young American people who kept running from the lab to the cold room: they
co Oxford University
Press

were Rich Losick and Jan Pero, who were spending a few months in P.Schaeffer's department and nervously checking the column collector which used to get stuck at the time of elution of their precious RNA polymerase fractions. Reading the literature was soon going to tell me who these people were but when I started working in the lab after the summer vacation they had gone back to the States and it would be many years before our paths crossed again. Part of the excitement about RNA polymerase in the early 1970s was due to the recent discovery of sigma, a protein factor binding to E. coli core RNA polymerase and conferring on it template specificity. In their seminal paper Burgess et al. (1969) had conjectured that several sigma factors could co-exist in bacteria (or in phage-infected bacteria) and activate specific sets of genes. Soon afterwards, the discovery that B. subtilis RNA polymerase lost its ability to transcribe phage 4be DNA when purified from sporulating cells led to the intriguing suggestion that the onset of sporulation in B. subtilis was accompanied by a change of template specificity of RNA polymerase which in turn led to the expression of a new class of genes (Losick and Sonenshein, 1969). Either a covalent modification of core RNA polymerase or the inactivation of vegetative sigma factor seemed to be required for such a change of template specificity in order to allow core RNA polymerase to bind a putative new sigma factor. Encouraged by the report of an alteration of the a subunits of core RNA polymerase after infection of E. coli by bacteriophage T4 (Goff and Weber, 1970), I decided to look for a similar phenomenon in sporulating B. subtilis. It was my turn to watch suspiciously over the column collector before analysing RNA polymerase fractions by in vitro transcription and SDS -polyacrylamide gel electrophoresis. Inspired by the T4 example I was specifically looking for the covalent binding of some phosphorous-containing compound but, after some transient hopes, I had to admit that core RNA polymerase was not phosphorylated during sporulation in B. subtilis. Meanwhile, some evidence had been provided that the vegetative sigma factor was inhibited early during sporulation (Tjian and Losick, 1974) and various polypeptides had been found that co-purified with core RNA polymerase from sporulating cells (Fukuda et al., 1975; Linn et al., 1975). I then chose to address a somewhat neglected problem, the molecular basis for the supposed compartmentalization of gene expression during sporulation. Isolation of forespores from Bacillus cereus and Bacillus megaterium had been reported (Andreoli et al., 1973; Ellar and Postgate, 1974) and I decided to follow similar approaches for comparing RNA polymerase composition in forespore and mother cell from B.subtilis. Purified RNA polymerase would then be used to transcribe nucleoids in vitro and the RNA products would be analysed in hybridization-competition experiments. My secret dream was to be able to purify a polypeptide with the characteristics of a forespore-specific sigma factor. All my efforts were then devoted to setting up a protocol for

3559

P.Stragier

++:t,' +
..

't
;

. ',
%,,
-fS'

;;
.
.. x
-.' [

',
._!
S d y
S

t:S

.:

sI +;i;g t>

,"*->

s;

_;

;4>> X

;|i ;8* 'tt '8 s!F i***_

- e X t a @ . - l l ~~~~~~~~~~~~~~~~~~~~ ... .r, ..~~~~~4r,

"r ._;xsX 8;<ow, '41''j s t X ' i'1 _ tAk W; q t. '' su -I~~~~~~~~~~~~~~~~~~~~~ d

f E,,,.,>,,,>;, S,SSM1

'

' d Xe,,g0S , Pf.~~ ~ ~ o

4'

Fig. 1. Electron micrograph of a sporulating B. subtilis cell at an intermediate to late stage of sporulation. The forespore appears as an organelle ( 0.5 -I rm) within the mother cell. The white area is the cortex, a peptidoglycan-like material, which is deposited between the two membranes of the forespore. The lamellar structures depositing on the outside of the forespore are the coats. Photograph kindly provided by A.Ryter.

purifying forespores in sufficient amounts and as free as possible from mother-cell material. The forespore fractions were examined by electron microscopy by Claude Frehel who was working in Antoinette Ryter's laboratory at the Pasteur Institute. Antoinette had been one of the pioneers who defined the morphological stages of sporulation in B.subtilis (Ryter et al., 1966) and I could not have found a better place for checking my samples. However, despite intensive and multiple attempts I was never able to obtain clean enough forespore fractions and Claude got used to greeting me with a slightly embarrassed smile before showing me her pictures of the latest experiment. I felt all the more frustrated since it had just been demonstrated by a very clever genetic approach that some sporulation genes were required only in the mother cell and others exclusively in the forespore (Lencastre and Piggot, 1979). This disappointment was also shared by many people who had started using the newly available cloning techniques to study B. subtilis. For many years it appeared almost impossible to propagate a recombined plasmid in B.subtilis without the insert being rearranged, which precluded any reliable complementation analysis. The remarkable recombination efficiency of B.subtilis was then a handicap which delayed the era of B. subtilis molecular genetics and led many people to switch to another field. Therefore the cloning of the first sporulation gene was a real tour deforce (Segall and Losick, 3560

1977). Many years would go by before other spo genes were cloned and they were a prerequisite for assaying the specificity of a putative sporulation sigma factor. It is ironic that the molecular genetics of B. subtilis are now so well developed, essentially as a result of the efficiency of homologous recombination: it is very easy to create mutations by reverse genetics, cloned DNA can be introduced back into the chromosome either at its original locus or at some other desired location and expression of any gene can be monitored throughout sporulation by fusing its promoter region to the E. coli lacZ gene. I was not in a position to wait for this golden age and, having to get a Ph.D. as soon as possible, I betrayed B.subtilis and moved to the other end of the corridor where I joined the lab of JeanClaude Patte who was working on regulation of lysine biosynthesis in E.coli.

A spo gene encodes a sigma factor The lysA gene, encoding the enzyme converting diaminopimelate (DAP) to lysine, had just been cloned and somebody had to sequence it. In those days sequencing was
still an adventure and I could not afford to lose more time. So I spent the summer of 1980 in Moshe Yaniv's lab at the Pasteur Institute where I learned how to play Lego with restriction enzymes and how to survive the Maxam and

Dances with sigmas

IP--

spoOH-I

Lri
0
germination starvation

I
2 chromosomes

t--II II
asymmetric

spoIIGB \ GE
tmw

spoIIAC cyF

septation

free spore

I
--

IE
#P
p-

III

PAM forespore
(aG) spoIIIG

'V

I
I
_ ~~~

01~-I
_ _ _

Am

,~~1-

mother
(aK)

cell

sigK
Fig. 2. The sporulation cycle. The various stages are indicated by roman numerals. The five sigma factors involved in sporulation names of their encoding genes at the time of their activation. Adapted from Losick et al. (1986).
are

indicated with the

Gilbert technique. Back in Orsay I dismantled the 4.4 kb fragment that I had inherited and found two genes in it, both required for lysA activity (Stragier et al., 1983b). One was lysA itself, and the second one, located upstream and in the opposite orientation of lysA, encoded a protein absolutely required for lysA transcription (Stragier et al., 1983a; Stragier and Patte, 1983). I called this regulatory gene lysR. I had no idea that this obscure LysR protein would later have so many cousins and become the paradigm for a family of bacterial regulators (Henikoff et al., 1988). Expression of biosynthetic pathways in bacteria is usually repressed by the final product of the pathway acting in conjunction with an aporepressor. In the case of amino acid biosynthesis the recent discovery of attenuation had shown that complex cis-acting sequences could induce premature transcription termination if the final product was present in excess (Lee and Yanofsky, 1977). Therefore it was quite a surprise to find that the final step of lysine biosynthesis in E.coli was controlled by an activator. This result could, however, be rationalized because DAP, the substrate of the lysA product, is also a major constituent of the bacterial cell wall. Since lysA expression was induced by DAP and repressed by lysine, the simplest interpretation was to assume that DAP and lysine modulated the activation of lysA transcription by LysR in opposite ways. In other words, building the cell wall had priority over making new proteins. That was my first contribution to regulation of transcription in bacteria and it allowed me to finally become a 'Doctor'. I was then seriously thinking about finding a post-doctoral position in the States and switching to some fancy eukaryotic system. But a conjunction of circumstances made me delay that project. First, several genes involved in DAP biosynthesis had now been cloned in the lab and none showed any attenuation-like sequence, which left the basis of their

regulation by lysine mysterious. Secondly, J.-C.Patte was moving to Marseilles, in the south of France, where he was going to work on Pseudomonas and, because of some administrative mix-up, no substitute teacher was appointed by the University to take up the lab. Third, I had enjoyed a cheerful and productive association with my colleague Jean Bouvier during the last year and we both wished to carry on for a while. So we were given some space and support to stay at the Microbiology Institute and to extend our investigations on the dap genes. We were joined by Catherine Richaud who had played a major role in cloning and analysing the genes of the DAP lysine pathway and who was now hunting for the last missing dap genes. We anticipated completing the whole story in a couple of years. Before he left for Marseilles in June 1983, J.-C.Patte gave a party at the Microbiology Institute and many of his colleagues were present. While I was enjoying the food and wine I was approached by Jekisiel Szulmajster. He was the last scientist in France to be still working on sporulation in B. subtilis. In his lab at the CNRS campus of Gif-sur-Yvette, a few kilometres from Orsay, Celine Bonamy had cloned the spoOB gene. He knew that I was a sequencing addict and he wanted me to sequence that piece of DNA. That seemed an easy way of settling an old score with B.subtilis and I agreed to do it with J.Bouvier. When we brought the spoOB sequence to Gif 3 months later, J.Szulmajster confessed that C.Bonamy had actually cloned a second gene, spoIIG, and suggested that we could also spend some time on its sequencing. We agreed but made it clear that there would be no third gene... After finishing some experiments on the expression of spoOB (Bouvier et al., 1984) and still working mainly on the dap genes, we sequenced spolIG in March 1984. I had kept my connections with M.Yaniv's lab and one of his students, Olivier Danos, advised me to
-

3561

P.Stragier

compare the sequence of the spoIIG product with all the available protein sequences that were stored in the Pasteur Institute computer in order to get some hint about its function. I had never heard of such an approach before and found the idea attractive. On April I 1 I introduced the spoIIG sequence into the Pasteur computer and ran the comparison program. And two hours later I had the answer: the spoIIG product contained a region which showed a highly significant similarity with part of the E.coli a70 factor! The similarity was such that it could be predicted that the spoIIG product was itself a sigma factor. During the preceding years the field of sigma factors in B.subtilis had moved quite fast. First, a new sigma factor, called a29, had been found to be present exclusively in sporulating cells (Haldenwang et al., 1981). Secondly, spoVG, the first sporulation gene to be cloned, was induced at the onset of sporulation and depended on minor vegetative sigma factors for its transcription in vitro (Johnson et al., 1983). Thirdly, the program of transcription after infection of B. subtilis with phage SPO 1 was controlled by phage-encoded sigma factors acting sequentially and recognizing highly conserved promoter sequences (Tjian and Pero, 1976; Talkington and Pero, 1979). Altogether these results had led to the proposal that sporulation was controlled by a cascade of sigma factors (Losick and Pero, 1981). This model was so attractive that it was almost immediately admitted as being correct and we had a hard time convincing the Editor of Nature that there was still no genetic evidence for it and that the spolIG sequence provided it (Stragier et al., 1984). Having found what appeared to be the first spo gene encoding a sigma factor after so many years of frustrating work on B. subtilis, I was not just going to forget all about it and I decided to share my time between E. coli and B.subtilis. The next question to address was the exact function of the spoIIG product. The obvious candidate was (X29 (which would soon be renamed aE) and with J.Bouvier I chose an in vivo approach. The idea was to express spoIIG in E. coli (since we were familiar with physiological experiments using that organism) and to monitor expression of a reporter gene fused to a aE-controlled promoter. Such a promoter was provided by Linc Sonenshein (whom I had known when he was a post-doctoral fellow in P.Schaeffer's lab) as the promoter of the spoIID gene which was efficiently recognized by e in vitro. We constructed the required plasmids and E. coli strains but could not find any induction of spoIID in the presence of the spoIIG product. A possible unexpected explanation was provided when we heard about the results of Bill Haldenwang, the discoverer of EJE, reporting that aE was synthesized as an inactive larger precursor, pro-uE (Trempy et al., 1985b). Therefore we started making a series of deletions in the proximal part of the spoIIG gene, selecting directly for the ability to induce expression of the spoIID promoter in E. coli. Many months would be necessary before we had truncated versions of the spolIG gene that showed JE activity in E. coli and, in the meantime, it was demonstrated by immunological analysis that our hypothesis was actually correct and that aE was the product of spoIIG (Trempy et al., 1985a).

the heat shock response in E. coli, had all the characteristics of a sigma factor in vitro (Grossman et al., 1984) and that it contained regions of high similarity with a70 (Landick et al., 1984). The sequence of ciA, the B.subtilis major vegetative sigma factor, also became available (Gitt et al., 1985) so it was then possible to analyse more thoroughly the conserved regions in these four sigma factors. With the expert help of Claude Parsot we found three regions of conservation. The most striking one, which was the one initially found in spolIG, covered about 80 amino acids and was located roughly in the first third of the proteins. We speculated that its proximal part was involved in binding of the sigmas to the core RNA polymerase, while its more variable distal region could modulate that interaction. The two other conserved regions appeared to contain potentially a similar secondary structure, the helix-turn-helix motif which was becoming well known as a putative DNA-binding domain. One of these regions was close to the carboxyterminal end of the sigmas and the other was internal. We proposed that these two regions were involved in recognition of the two highly conserved sequences found in bacterial promoters and reported these hypotheses in a paper that was very difficult to get published because of the absence of any experimental data (Stragier et al., 1985). Since then, biochemical and genetic evidence have been provided that support part of our model: the most conserved intemal region is actually involved in binding to core RNA polymerase (Lesley and Burgess, 1989) and the carboxy-terminal helix-turn-helix motif interacts with DNA sequences located 35 bp upstream of the transcription start (Gardella et al., 1989; Siegele et al., 1989). No function has been found for the other putative helix -turn -helix region and recognition of DNA sequences located 10 bp upstream of the transcription start is provided by the more variable region adjacent to the core-binding domain (Siegele et al., 1989; Zuber et al., 1989). Today 30 sequences of bacterial sigma factors are known which fit the modular scheme shown in Figure 3, while a few others (the 'aJ54 family' (Kustu et al., 1989) and the phage-encoded sigmas) have a completely different organization. I was very confident in our model and I was eagerly waiting for other sigma sequences that would confirm it. So, when the rumour came by mid-1985 that another spo gene, spolIAC, also encoded a sporulation sigma factor (Errington et al., 1985) I feverishly studied its sequence that had been published a few months before without any mention of a putative function of the spoIIAC product because comparing a new sequence with others in the database was not yet normal practice. The spoIIAC sequence aligned very well with the other sigmas, but it stopped before the conserved carboxy-terminal domain. I could not believe it and since the data did not fit the model, the data had to be incorrect! Examining closely the nucleotide sequence of spoIIAC, I found that the missing similarity was actually present, but downstream of the stop codon and in another reading frame.

NH2

CORE -KCO2H

Conserved domains in sigma factors

Just a few weeks after we obtained the spoIIG sequence it was reported that the product of htpR, the gene controlling

organization.

Fig. 3. Schematic structure of bacterial sigma factors. Boxes indicate regions of sigma factors involved in binding to core RNA polymerase or to the - 10 and -35 regions of promoters. The sigma factors belonging to the 4 family and the bacteriophage-encoded sigma factors have a different

3562

Dances with sigmas

It was just a matter of adding one nucleotide here and one there and the spoIIAC product would be quite a decent sigma factor. This was diplomatically suggested to the authors and the missing nucleotides were found where they had to be. Then I knew with certitude that our alignments of sigmas conveyed powerful information that could be very useful if correctly interpreted (Stragier, 1986). And 3 years later these alignments would give me an astonishing reward.

Controlling the activation of a sigma factor

The identification of the spolIG product as aE left unexplained the significance of the synthesis of oE as an inactive precursor pro-oE. Physiological and genetic experiments had to be done in B.subtilis and my own experience was limited to growing bacteria before breaking them to prepare an extract. I desperately needed someone with a good knowledge of B.subtilis and I found the right person on the next floor, Celine Karmazyn-Campelli, who was a former student of P.Schaeffer. A spoIID-lacZ fusion was introduced into B. subtilis and could be induced during vegetative growth by a truncated form of aE that was under the control of an IPTG-inducible promoter. Conversely, the intact spoIIG gene was inefficient during growth but worked very well if induced after the onset of sporulation. Interestingly, the amino-terminal part of pro-aE showed the ability to form an amphiphilic helix and, by analogy with the targeting sequences of imported mitochondrial proteins, I fancied the idea that the aE pro-sequence could play a similar role and that pro-oE processing took place while the pro-rE molecules were crossing the septum double membrane. Such a mechanism could be used to segregate the active aE molecules in one of the two compartments of the cell. According to that model the pro-oE processing machinery was expected to contain transmembrane and cytoplasmic domains (or subunits), playing the role of the pore and of the peptidase respectively. There was an excellent candidate for the pro-oE processing enzyme and that was the product(s) of the spoIIE locus. This was the only locus known in which mutations did not interfere with pro-oE synthesis but prevented its conversion to rE (Trempy et al., 1985b). Internal fragments of spollE were provided to us by Mike Young and we started cloning the remaining part of this locus as a first step towards sequencing it. Having learned that sequencing of spoIIE was already well advanced in Phil Youngman's laboratory, we focused our efforts on studying spollE expression in relation to spolIG. Both loci were found to be expressed at the same time, about one hour before spolID which was used as an indicator of aE activity and therefore of pro-oE processing. Thus the situation was quite confused by mid-1986 when C.Bonamy decided to spend one year in my laboratory. She had found that one of the classical spolG mutations was not in the gene encoding pro-oE but in another gene located immediately upstream and I thought that it could be informative to sequence this other gene. The spolIG locus was found to contain two genes organized in an operon and, following the usual nomenclature, we called the first one spoIIGA and the second one, which encoded pro-orE, spoIIGB. It was quite a surprise when I realized in October 1986 that the sequence of the spoIlGA product had all the features that I had predicted for the pro-uE processing enzyme and it took only one night to decide that everybody was wrong

about the role of spoIIE and that we had the true pro-aE processing enzyme in our hands. The following months were frantic. I did not want to publish the spoIIGA sequence without providing some evidence about its function. An in-frame deletion was created in spoIIGA and found to block spoIID transcription, suggesting that pro-uE was not processed in that mutant. More direct proof was obtained by constructing a gene encoding an hybrid pro-aE-3-galactosidase protein which could be partially processed in a wild type strain but not in the spoIIGA mutant. There seemed to be no doubt that SpoIlGA was required for pro-uE processing. But was it sufficient? To answer that question we induced the synthesis of both SpoIIGA and pro-uE during vegetative growth and found a significant induction of spoIID-lacZ. Therefore we concluded that the spoIIGA product was likely to be the proaE processing enzyme. But how could we explain why proaE accumulated for about one hour before getting processed if the processing enzyme was synthesized simultaneously? And what was the role of the spoIIE product(s)? The situation was even more complicated because mutations in the spoIIA operon (which encoded the other putative sporulation sigma factor soon to be known as oF) also blocked spoIID-lacZ transcription, although both the spoIIG and the spoIIE operon were normally expressed. In fact too many gene products
S p1., I IA A
-

F-

r ,eptum1

rn
S

LL (mpcnen[N

p0 I I1

AL rl /IB
'

AAxA ...N"r

OREP

I.I

M(i

K (''-( (1ili R

E-13

Fig. 4. A model for activation of aE. The SpoIlGA protein is shown as a box partly embedded in the membranes of the sporulation septum (stippled area). Light arrows indicate the involvement of accessory gene products in pro-aE processing (symbolized by the thick arrows). Part A is adapted from Stragier et al. (1988) when it was believed that aE and aF were active in both compartments. Part B is an updated version where pro-oE processing takes place only in the mother cell in response to a cascade of interactions leading to activation of aF in the forespore.

3563

P.Stragier

were required to get active oE in the sporulating cell! It was finally possible to build a coherent model from all these data by assuming that SpoIIGA became active only when inserted in the septum membrane, because of some specific properties which required the spoIIE product(s) and the products of some other genes controlled by (F. The existing data suggested that spoIID was transcribed in both compartments and it was then concluded that activation of SpoIIGA occurred on both sides of the septum (Figure 4A). In this model, activation of the aE-controlled regulon was strictly dependent on the presence of the sporulation asymmetric septum and gene expression at an intermediate stage of sporulation appeared to be coupled to the successful completion of a previous morphological step (Stragier et al., 1988). Very recently new data have been obtained which strongly indicate that fF becomes active only in the forespore compartment in response to a cascade of interactions initiated by the spolIE products (Margolis et al., 1991) and that fE becomes active only in the mother cell, suggesting that pro-urE processing takes place only in that compartment (Driks and Losick, 1991). An updated model is shown in Figure 4B where the septum still plays a crucial role in the vectorial transfer of activation from aF to (E.

the forespore and are degraded during germination (Setlow, 1988). The ssp genes encoding the SASPs had been cloned in P.Setlow's lab and had been used as in vitro templates to purify a factor allowing core RNA polymerase to recognize their promoters. P.Setlow was telling R.Losick that an amino-terminal sequence of the factor had been determined but it did not match any known spo gene. R.Losick's reaction was quick and a few minutes later I was on the phone dictating the amino-terminal sequence of a0G and getting enthusiastic approvals from P.Setlow: the aG sequence matched exactly the sequence of his putative forespore-specific sigma factor ! Afterwards it was just a matter of weeks before we could demonstrate that synthesis of a0 during vegetative growth induced expression of all known forespore-specific genes (Sun et al., 1989). The function of oG being unravelled C.Karmazyn-Campelli could concentrate on the regulation of its synthesis and she was to find multiple and complex levels of controls (Karmazyn-Campelli et al., 1989; Stragier, 1991). It appears now that o is itself synthesized exclusively in the forespore at stage II, but becomes active only later at stage III, in response to the end of engulfment.

The forespore sigma factor

The sequence of the spolIG locus showed the presence of an incomplete open reading frame located only 140 bp downstream of spoIIGB. Could it be the third gene of the spolIG operon? By mid-1987, in order to address this question, we engineered a disruption of this gene and found that the mutated cells were asporogenous, their sporulation being arrested at stage IH (see Figure 2). Therefore the open reading frame downstream of the spoIIG operon defined a new sporulation gene which we called spoIIIG. On June 29, looking with a renewed interest at the partial sequence of the spoIIIG product, I suddenly recognized motifs that had been posted above my desk for the past 3 years: this was a sigma! And a very interesting one since it was required at a stage of sporulation where gene expression was known to be compartmentalized. This sigma, which we called o, could be forespore or mother-cell specific. Since only part of the spoIIG gene was present in the insert cloned by C.Bonamy we 'walked' on the B.subtilis chromosome to clone the missing part and the sequence was completed by the end of October. Moreover, we had a plasmid carrying the complete spoIIIG sequence under the control of an IPTGinducible promoter and I could now address the question of the physiological role of aG. During all these years we had been handicapped by the poor communication with our colleagues on the other side of the Atlantic and to change that situation I decided to spend one month in R.Losick's lab at Harvard University. I wanted to look for genes controlled by aE and aG by screening a library of transposon-generated lacZ insertions using our plasmids in which rE and aG synthesis was IPTG inducible. Alan Grossman was using the same strategy for identifying genes controlled by uH, a recently identified sigma factor involved in the transition to post-exponential phase and I would greatly benefit from his advice. Thus, in November 1987, I was carefully looking at my X-gal agar plates when R.Losick got a phone call from Peter Setlow, the world authority on SASPs, the small proteins that accumulate in

sigma factor in pieces

After such an exciting visit to Harvard I decided to come back one year later. I had no special experiments in mind but I knew it was a good time to be there: Lee Kroos had purified a sigma factor (to be called aK) which allowed RNA polymerase to transcribe in vitro genes expressed at a late stage in the mother cell. From its amino-terminal sequence an oligonucleotide probe had been designed and a putative sigK gene had been cloned. On October 12 Barbara Kunkel showed me her first sequence of sigK which contained an excellent match with the conserved core binding region. Looking further up on the autoradiogram it was possible to guess the region corresponding to the oligonucleotide probe, but no typical pyrimidine-rich region that would be the ribosome binding site on the complementary strand could be seen immediately upstream. This suggested that aK might also be synthesized as an inactive larger precursor, pro-aK. We were then encouraged to look back to all the available sequences of spo genes and to search for the presence of the aminoterminal sequence of aK in an internal position. And we soon found it in spoIVCB, a gene known to be expressed only in the mother cell and from which only the first 40 codons had been sequenced. But there was still a mystery: the coding sequence for aK was interrupted by a stop codon at what should have been only half of the sigma and the similarity with the other sigma factors could not be restored by shifting the reading frame and reading further downstream. This was the symmetrical situation to that reported for the spoIIIC gene, which had been found to encode a product showing similarity to the carboxy-terminal part of sigma factors (Errington, 1987). And spoIlC was quite close to spoIVCB on the chromosome... It was then tempting to imagine that the two genes spoIVCB and spoIIIC were brought together by a DNA rearrangement that would create the intact sigK gene in the mother cell. A similar phenomenon was known to occur in some nitrogen-fixing bacteria during the differentiation process that leads to the activation of the nif operons (Golden et al., 1985). We were

3564

Dances with sigmas

a
__

..._ _...

nutrient starvation

transcriptional
induction
H

post-

%,r-

asymmeticn
septation
inactivation of SpoIIAB in the forespore
aF

SpoIIAA-dependent

SpoIIGA-mediated pro-GE processing

b

in the mother cell

4(YE 7
end of

engulfment
C_A

SpolIIA-dependent
activation of aG in the forespore

G
-..........

cortex

synthesis
Fig. 5. The chromosomal rearrangement creating sigK. The sigK gene is composite of the spoIVCB and spoIlC genes which are joined in frame by excision of an intervening sequence. The rearrangement occurs only in the mother cell and is dependent on the spoIVCA product. Reproduced with permission from Kunkel (1991).
a

forespore-dependent pro-aK processing

in the mother cell

coat

synthesis

4K aK

all excited about this possibility and R.Losick agreed that I should start immediately with B.Kunkel the experiments that would assess that specific hypothesis. DNA was prepared from vegetative cells as well as from late sporulating cells and digested with various restriction enzymes. The physical map of the spoIVCB and of the spollIC regions was then determined by Southern blotting using specific probes for each of the two genes. And 10 days after our first examination of the sigK sequencing gel we had an autoradiogram demonstrating that the two loci were actually rearranged during sporulation. During the following week the rearranged gene was cloned from sporulating cells and I had one of my best birthday presents when I read the sequence across the rearrangement breakpoint which confirmed all our hypotheses (Stragier et al., 1989). In the following months these experiments were continued on both sides of the Atlantic. The rearrangement was found to be due to the excision as a closed circle of an intervening sequence of about 42 kb and to occur only in the mother cell. This was the expected result since the mother cell compartment is lost at the end of sporulation while the intervening element has to be still present in the spore to be transmitted to the next generation of bacteria. Interestingly the rearrangement is controlled by the spoIVCA gene which is carried by the element itself (Figure 5). At this stage of the results the rearrangement was considered to be a key element in establishing compartmentalized gene expression in the mother cell by creating the gene for the mother cell specific sigma factor aK. However, it was possible to

Fig. 6. Cascade activation of sporulation sigma factors. Specific requirements for activation of each sigma factor are boxed and depend themselves on the action of a previous sigma factor. The morphological stages of sporulation are indicated at their approximate position according to the transcriptional cascade. Adapted from Stragier and Losick (1990) where the role of the various spo genes involved in the cascade is detailed. construct a strain in which the interrupted sigK region
was

replaced by an intact rearranged sigK gene. That strain grew and sporulated normally although a rearranged copy of the sigK gene was present in the forespore (Kunkel et al., 1990). Therefore the rearrangement per se is not required for allowing aK synthesis exclusively in the mother cell. Restriction of transcription of sigK and of processing of proUK to the mother cell are the actual mechanisms that establish mother cell specific gene expression (reviewed by Kunkel, 1991).

new

challenge

Since 1987 I had definitely quit E. coli and the study of the dap genes and I had progressively attracted new people to work on sporulation of B.subtilis. So, when Marianne Grunberg-Manago offered me a spacious lab in her department at the IBPC, I did not hesitate and in April 1989 my lab moved to Paris. Since then other avenues of sporulation have been explored and the sigma factor saga has expanded. It is now clear that for all the five sigma factors known to be required during sporulation (as indicated in Figure 2) the ultimate and most important control is exerted at the level of their activity. The cascade of sigma factors predicted 10 years ago is a cascade of sequential
3565

P.Stragier

to pay me back for all the lost years of my youth and she taught me little by little how to tame the mythical sigmas. They became familiar inhabitants of the landscape along the receding frontier where I stayed posted. Now I want to go across and to explore a new world where cells talk to each other and secret messages are buried in morphological structures. It should also be fun. During all these years I have had the privilege to work with many good friends and colleagues, among whom I have a special debt to Jean Bouvier, Celine Karmazyn-Campelli and Rich Losick. I am grateful to David Popham for his help during the preparation of this manuscript.

grasp all the opportunities. Maybe Mother Nature decided

activations in which each sigma depends upon a previous one for becoming active as summarized in Figure 6 (Stragier and Losick, 1990). Most interestingly, the major part of the cascade is compartmentalized and there appears to be a conversation between the forespore and the mother cell, each compartment telling the other one when to proceed to the next developmental step by sending signals conveyed through the completion of specific morphological structures. As I tried to make clear in this review I went myself through a cascade of bits of luck and, thanks to the freedom given by CNRS and my successive advisors, I was able to

Losick,R., Youngman,P. and Piggot,P.J. (1986) Annu. Rev. Genet., 20, 625-669. Margolis,P., Driks,A. and Losick,R. (1991) Science, in press. Ryter,A., Schaeffer,P. and Ionesco,H. (1966) Ann. Inst. Pasteur Paris, 110, 305-315. Segall,J. and Losick,R. (1977) Cell, 11, 751-761. Setlow,P. (1988) Annu. Rev. Microbiol., 42, 319-338. Siegele,D.A., Hu,J.C., Walter,W.A. and Gross,C.A. (1989) J. Mol. Biol., 206, 591-603.

Stragier,P. (1986) FEBS Lett., 195, 9 -1 1. Stragier,P. (1991) In Cole,J.A., Mohan,F. and Dow,C. (eds), Procaryotic Structure and Function. Society for General Microbiology, in press. Stragier,P. and Losick,R. (1990) Mol. Microbiol., 4, 1801 -1806. Stragier,P. and Patte,J.-C. (1983) J. Mol. Biol., 168, 333-350. Stragier,P., Danos,O. and Patte,J.-C. (1983a)J. Mol. Biol., 168, 321-331. Stragier,P., Richaud,F., Borne,F. and Patte,J.-C. (1983b) J. Mol. Biol., 168, 307-320. Stragier,P., Bouvier,J., Bonamy,C. and Szulmajster,J. (1984) Nature, 312, 376-378. Stragier,P., Parsot,C. and Bouvier,J. (1985) FEBS Lett., 187, 11-15. Stragier,P., Bonamy,C. and Karmazyn-Campelli,C. (1988) Cell, 52, 697-704. Stragier,P., Kunkel,B., Kroos,L. and Losick,R. (1989) Science, 243,
507-512.

Sun,D., Stragier,P. and Setlow,P. (1989) Genes Dev., 3, 141-149. Talkington,C. and Pero,J. (1979) Proc. Natl. Acad. Sci. USA, 76,
5465 -5469.

References
Andreoli,A.J., Suehiro,S., Sakiyama,D., Takemoto,J., Vivanco,J.C. and Klute,M.C. (1973)J. Bacteriol., 115, 1159-1166. Bouvier,J., Stragier,P., Bonamy,C. and Szulmajster,J. (1984) Proc. Nati. Acad. Sci. USA, 81, 7102-7106. Burgess,R.R., Travers,A.A., Dunn,J.J. and Bautz,E.K. (1969) Nature, 221, 43-46. Driks,A. and Losick,R. (1991) Proc. Natl. Acad. Sci. USA, in press. Ellar,D.J. and Postgate,J.A. (1974) In Barker,A.N., Gould,G.W. and Wolf,J. (eds), Spore Research 1973. Academic Press, London and New

Tjian,R. and Losick,R. (1974) Proc. Natl. Acad. Sci. USA, 71, 2872-2876. Tjian,R. and Pero,J. (1976) Nature, 262, 753-757. Trempy,J.E., Bonamy,C., Szulmajster,J. and Haldenwang,W.G. (1985a) Proc. Natl. Acad. Sci. USA, 82, 4189-4192. Trempy,J.E., Morrison-Plummer,J. and Haldenwang,W. (1985b) J. Bacteriol., 161, 340-346. Zuber,P, Healy,J., Carter,H.L.III, Cutting,S., Moran,C.P.Jr and Losick,R. (1989) J. Mol. Biol., 206, 605-614.

York, pp. 125-166. Errington,J. (1987) FEBS Lett., 224, 257-260. Errington,J., Fort,P. and Mandelstam,J. (1985) FEBSLett., 188, 184-188. Fukuda,R., Keilman,G., McVey,E. and Doi,R.H. (1975) In Gerhardt,P., Costilow,R.N. and Sadoff,H. (eds), Spores VI. American Society for Microbiology, Washington, DC, pp. 213 -220. Gardella,T., Moyle,H. and Susskind,M.M. (1989) J. Mol. Biol., 206, 579-590. Gitt,M.A., Wang,L. and Doi,R.H. (1985) J. Biol. Chem., 260, 7175-7178. Goff,C.G. and Weber,K. (1970) Cold Spring Harbor Symp. Quant. Biol., 35, 101-108. Golden,J.W., Robinson,S.J. and Haselkorn,R. (1985) Nature, 314, 419-423. Grossman,A.D., Erickson,J.W. and Gross,C.A. (1984) Cell, 38, 383-390. Haldenwang,W.G., Lang,N. and Losick,R. (1981) Cell, 23, 615-624. Henikoff,S., Haughn,G.W., Calvo,J.M. and Wallace,J.C. (1988) Proc. Natl. Acad. Sci. USA, 85, 6602-6606. Johnson,W.C., Moran,C.P.Jr and Losick,R. (1983) Nature, 302, 800-804. Karanyn-Campelli,C., Bonamy,C., Savelli,B. and Stragier,P. (1989) Genes Dev., 3, 150-157. Kunkel,B. (1991) Trends Genet., 7, 167- 172. Kunkel,B., Losick,R. and Stragier,P. (1990) Genes Dev., 4, 525-535. Kustu,S.,E. Santero,E., Keener,J., Popham,D. and Weiss,D. (1989) Microbiol. Rev., 53, 367-376. Landick,R., Vaughn,V., Lau,E.T., VanBogeler,R.A., Erickson,J.W. and Neidhart,F.C. (1984) Cell, 38, 175-182. Lee,F. and Yanofsky,C. (1977) Proc. Natl. Acad. Sci. USA, 74, 4365-4369. Lencastre,H. and Piggot,P. (1979) J. Gen. Microbiol., 114, 377-389. Lesley,S.A. and Burgess,R.R. (1989) Biochemistr, 28, 7728-7734. Linn,T., Greenleaf,A.L. and Losick,R. (1975) J. Biol. Chem., 250, 9256-9261. Losick,R. and Pero,J. (1981) Cell, 25, 582-584. Losick,R. and Sonenshein,A.L. (1969) Nature, 224, 35-37.

3566