and single-strand-specific cleavage. The two open com- plexes produce identical deoxyribonuclease I footprints and similar methylation interference/protection ( ...
196
RNAP
BIOCHEMICAL SOCIETY TRANSACTIONS
+ DNA G Closed
-
0,
7
Short RNA
-Long
RNA
Abortive transcript
Shon RNA
-
Open complexes
Long RNA
Initiated comDlexes
Fig. 1. Summary of the intermediates in transcription initiation
Abbreviation: RNAP, RNA polymerase.
and single-strand-specific cleavage. The two open complexes produce identical deoxyribonuclease I footprints and similar methylation interference/protection (one base shows different behaviour), indicating that both complexes are at the same binding site. The two open complexes, however, differ in the amount of unwinding in the - 13 to 3 region, as assayed by probes for singlestranded DNA. A functional assay, transcription within the gel slices containing the open complexes, indicates that the open complexes also differ in the ability to escape from the abortive initiation cycle into productive transcription. The upper gel band 0, has a lower total RNA synthesis rate but a greater relative production of the stably bound 11-mer compared with the abortive 8-mer product. Dominance of 0, at higher temperature may reflect an adaptive mechanism to counter the observed tendency of abortive
+
initiation to increase with increasing temperature. The results are summarized in the scheme shown in Fig. 1. Deoxyribonuclease I footprinting of the initiated complex containing only the first stable transcript (the 1 1-mer) shows a contraction from the open-complex footprints by loss of approximately 25 bp of contacts in the - 35 region, and loss of the o-subunit. This drastic structural change, and the measured apparent activation energy for abortive release of nascent RNA transcripts, lead us to propose a model which features an energetically stressed intermediate in the translocation of polymerase away from its contact in the - 35 region. The model focuses on mutually competitive interactions of polymerase with the - 35 region or with short transcripts in the 0 to 10 region; the stability of the RNA-DNA hybrid double helix formed by the pascent transcripts can also be expected to affect the relative importance of abortive initiation.
+
The organizers thank Gibco/BRL for their sponsorship of this lecture.
Beckwith, J. & Zipser, D. (1970) The Lactose Operon, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY Chamberlin, M. J. (1974) Annu. Rev. Biochem. 43, 721-775 Fried, M. & Crothers, D. M. (1984~)J . Mol. Biol. 172, 241-262 Fried, M. & Crothers, D. M. 19846) J . Mol. Biol. 172, 263-282 McClure, W. R., Cech, C. L. & Johnson, D. E. (1980) Proc. Natl. Acad. Sci. U.S.A. 77, 5634-5638 Reznikoff, W. S. & Abelson, J. (1978) in The Operon (Miller, J. H. & Reznikogg, W. S., eds.), pp. 221-243, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY Siebenlist, U. (1979) Nature (London) 279, 651452 Wu, H. M. & Crothers, D. M. (1984) Nature (London) 308,509-513
Mechanism of activation of transcription by the complex formed between cyclic AMP and its receptor in Escherichia coli H . BUC Institut Pasteur, 25, rue du Docteur Roux, 75724 Paris Ckdex 15, France Our present knowledge about proteins acting on DNA relies mainly on static descriptions of their mutual interactions. The picture of a DNA-protein complex as deduced from X-ray crystallography yields clues about its function, in particular when the role of the protein, in the cell, is itself rather static. However, when the regulatory action of a protein is exerted through the control of a rate, this description is generally insufficient if not misleading. No punter will bet on a horse, just because he likes its picture at rest. Escherichia coli RNA polymerase is a molecular device optimized for the control of rates. Initiation startpoints which are positively controlled in the bacterial cell are generally weak promoters (McClure, 1985; Raibaud & Schwartz, 1984). Specific defects have been introduced during the course of evolution at defined loci in the primary sequence of such promoters. They can be selectively corrected by the presence of an activator. In order to understand such processes, it is essential to know, first, which structural changes are usually associated with the various steps leading to open-complex formation, second, which steps are impaired when the promoter sequence is specifically altered by ‘down’ mutations, and, last, how an activator protein acts as a catalyst to overcome the specific defect at the relevant kinetic step. We have undertaken a systematic comparison of the kinetics of open-complex formation at the lac, gal and Abbreviation used: CRP, cyclic AMP-binding gene-activatingprotein.
malT control regions of E. coli. All these promoters are positively controlled by the complex formed between cyclic AMP and its receptor protein, CRP. In the case of lac, the wild-type promoter is compared with the wellknown ‘up’ mutant lac UV5, which is hardly activated by cyclic AMP-CRP either in vivo or in vitro (cf. de Crombrugghe et al., 1984). Three major diflerences in the pathway leading to opencomplex formation at lac UV5 and at lac wild-type Firstly, the reactivity of the Pribnow box in the lac wild-type lac ps or lac UV5 promoters are different towards various nucleolytic reagents (Sigman et al., 1985; A. Spassky & D. Sigman, unpublished work). The differential pattern is rather extended, suggesting a change in the structure of the DNA itself. Second, at lac UV5, a single type of open complex and a single set of 1 are observed (the + 1 transcripts starting around start is the physiological one): we say that the occupancy of the P1 promoter is 1. In contrast, on linear DNA fragments or on templates of moderate degree of supercoiling, the wild-type lac promoter yields two types of transcripts SI and S2 (starting at I and - 2) and two types of open complexes of rather long residence time (Maquat & Reznikoff, 1978; Malan & McClure, 1984; Spassky et al., 1984). The muted Pribnow box in the wild-type sequence does not direct RNA polymerase towards the efficient P1 promoter, and significant binding at a competing promoter, P2, is now possible. Third, the unwinding at both promoters does not take place in the same way as the reaction leading to open-complex formation proceeds. For this analysis a simple scheme is
+
+
I986
614th MEETING. OXFORD
197
used for both the P1 and P2 pathways (Chamberlin, 1974). R R
+ P’
ki
RP: A RP; k;
Kg.
kf“
+ PNeRPe-RP;
kr
KL
The measurement of the total unwinding generated by RNA polymerase at lac PI (for a lac UV5 insert) or at lac P2 (for a lac wild-type insert) has been performed by Amouyal & Buc (1986; cf. also Wang et al., 1977; Kolb & Buc, 1982). A value of - 17 f 0.3 turn has been found in both cases. On the other hand, on and off rate constants have been found in both cases. On the other hand, on and off rate constants have been measured on templates of increasing superhelical densities at lac UV5 (Malan et al., 1984) and at lac wild-type (Malan et al., 1984; Amouyal & Buc, 1986). These measurements provide semi-quantitative estimates of the amount of unwinding taking place at each step of the pathway provided that several controls and corrections are made. At lac UV5 a large unwinding occurs already during closed-complex formation (A0 = - 1.3 f 0.3 turn). Passage from RP, to the transition state located between RP, and RP,, RP* is associated with a significant rewinding (+0.3 f 0.1 turn); during the formation of the open complex at the expense of RP*, a small unwinding of about - 0.2 turn occurs. In constrast, for the wild-type lac promoter no significant unwinding takes place during formation of the initial closed complexes. It occurs in part at the transition state, in part after the rate-limiting step. Therefore, at lac UV5, the topology of the DNA is stepwise affected. Closed complexes, formed before the rate-limiting step, are already partially unwound. These intermediates commit RNA polymerase to the formation of the proper active (open) complex. The final unwinding, measured at lac wild-type, is as large as that observed in the ‘up’ mutant. (This observation shows, once more, that static measurements are of no great help understanding the influence of negative supercoiling on a dynamic process.) However, the reactivity of a wild-type promoter with respect to several nucleases is different and the closed complexes which are formed in this case are not topologically unwound: they generally lead to the formation of an incorrect complex of long residence time. Stable complexes at lac UV5: unwinding assists singlestranded formation At lac UV5, we found, with McClure, that the simple scheme outlined above was insufficient to account for out kinetic measurements (Buc & McClure, 1985; B. Mulligan & W. R. McClure, unpublished work). A sequential scheme involving two closed intermediates was found to be adequate. A similar model was proposed for the T7A, promoter (Kadesch et al., 1982; Rosenberg et al., 1982) and for the lamba P, promoter (Roe et al., 1984).
R
+ P =RP, KB
k2
c.RP, k-2
k3
=RP, k-3
For the lac UV5 promoter, it has been possible to give a preliminary characterization of the state of the DNA in the three binary complexes (Spassky et al., 1985; cf. also Kirkegaard et al., 1983; Becker & Wang, 1984 for the RP, complex, as well as Chenchick el al., 1981; Siebenlist et al. 1980 and references therein for the RP, complex). The rate-limiting step (in vivo and in most conditions, in Vol. 14
vitro) corresponds to the passage from RP, to RP, . Both complexes are closed: no significant single-stranded regions have been detected in either of these species. Hence open-complex formation is not limited by the formation of an open region in the promoter DNA. Formation of single strands takes place after the rate-limiting step, during the last isomerization which corresponds to a very large induced-fit. At the crucial transition state RP*, located between RP, and RP,, the DNA is topologically unwound but not yet single-stranded. At this stage, the DNA must be adequately bent or twisted to allow the rapid formation of a single-stranded region later on. The two complexes, RP, and RP,, correspond to two alternate modes of binding of the enzyme at lac UV5. RP, is stable at low temperatures (around 14°C) and RP, at 37°C (cf. Spassky et al., 1985). When the temperature is lowered, the positioning of RNA polymerase with respect to the DNA backbone, as probed by nucleolytic reagents, remains essentially unchanged. On the contrary, the pattern of differential reactivity of specific base residues towards methylation which is clear in RP, is lost in RP, . Furthermore, the single-stranded region extending between - 10 and + 3, re-anneals in the same narrow temperature interval. Interconversion between these two forms is fast at 37°C. The transition is co-operative and driven by entropy. Furthermore, as judged from the topological measurements quoted above, the formation of the single-stranded region, 13 base-pairs long, is accompanied by a very modest change in linking number. I conclude that the topological unwinding of - 1 turn already present in RP, is used at this stage to generate strand separation. A similar suggestion was made for the mechanism of open-complex formation at the tyrT promoter (Drew et al., 1985; A. Travers, personal communication). All these features indicate that an early defect in linking number is used to assist the formation of the functional single-stranded region. Unwinding in closed complexes might correspond in part to D N A wrapping
A process other than strand separation must contribute to the large deficit in linking number occurring in the first steps of the kinetic pathway. I re-examined the data in the literature to see whether this could not be partially accounted for by the wrapping of the DNA around part of the protein to form a negative superhelix. Obviously, some mild untwisting of the double helix could also take place at this stage, for example to put the - 35 and the - 10 region in register (cf. Stefan0 & Gralla, 1982; Buc & McClure, 1985). Two lines of evidence suggest the possibility of such a wrapping. First, when the temperature is lowered below 14”C, a series of inactive complexes is observed and T. Kovacic has observed in J. Wang’s laboratory that the pattern of nuclease attack displayed by these complexes is consistent with a structure having a periodicity of 1&11 base-pairs (as in the gyrase DNA complex studied by Kirkegaard & Wang, 1980). This pattern partially persists at 14°C or 37°C (cf. Spassky et al., 1984, 1985, for example). Second, Chenchick el al. (1981) have idenitfied by covalent cross-linking a rather unusual pattern of interaction between lac UV5 promoter DNA and RNA polymerase in the open complex at 37°C. Both DNA strands have multiple contacts with the three subunits /?, /?’and 0 . Six major areas of interaction between protein and DNA have been identified by these authors, each subunit contacting the promoter at least twice. These patches are interspersed. I assume that each ‘patch’ corresponds to the contact of a particular globular
BIOCHEMICAL SOCIETY TRANSACTIONS domain of a protein subunit, the interaction between the different globules (in the stable complexes) helping to maintain upstream and downstream regions of the promoter in a strict spatial relation. If the subunits are arranged linearly on the DNA, as proposed by Chenchick et al. (1981), the interspersed character of the area of contact poses a problem. This is avoided if one imagines that each subunit is equivalent to a step in a staircase contacting the DNA at its two different ends. The DNA is then wrapped around the protein contacting each subunit twice. This staircase model is reminiscent of the picture emerging from the more recent data on the nucleosome (cf. Richmond et al., 1986). The similarity can be extended if the DNA is wrapped in a lefthand superhelix around this protein core. This can be accomplished by tilting the steps with respect to the axis of the staircase. Such an operation generates a locally negative writhing. The sum of the untwisting required to put all the contacts on the same side of the DNA helix and of this writhing could easily account for the total topological unwinding in the closed complexes (roughly - 1.O turn) or in the open complex if part of the writhing is lost in favour of the formation of a single-stranded region, as assumed above. On the action of the cyclic AMP-CRP complex at lac UV5, lac wild-type and gal promoters The mode of action of the cyclic AMP-CRP complex has been examined using the abortive initiation assay. At lac UV5, the complex favours the formation of the closed complex RP, by increasing K,. At lac wild-type its action is dual. It prevents RNA polymerase engaging in the less efficient pathway leading to P2. Concommitantly, it increases KB(as for lac UV5) without affecting much the other rate constants (Malan et al., 1984). The gal control region is also composed of two overlapping promoters; they are known to be functional both in vivo and in vitro. Here again it has been shown that the action of the complex is dual. It shuts off the P2 pathway and activates the formation of open complexes corresponding to P1 (cf. Spassky et al., 1984). The gal control region provides an interesting example of two overlapping promoters which compete for RNA polymerase. A thorough kinetic analysis of this case has been performed by M. Herbert, A. Kolb and myself. Two relaxation rates govern the formation of open complexes at P1 and at P2. In the first step, the two open complexes are sorted out according to their rates of formation. (This is thought to be the only step of physiological significance.) In the last step a slow re-equilibration occurs during which the P1 promoter is favoured over P2 because it is more stable. This study demonstrates that the two paths leading respectively to P1 and to P2 are coupled: isomerization from the open complex formed at P2 to the open complex formed at P1 must proceed through the relevant closed complexes RP; and RPZ . The action of the cyclic AMP-CRP complex on this coupled pathway is difficult to understand directly. Promoter mutants selected by S. Busby and H. Fritz, and blocked either in the P1 pathway or in the P2 pathway, can be used in a way rather similar to the UV5 mutant in the lac case. Our present data indicate that here again the complex formed between cyclic AMP and CRP prevents the formation of the closed complexes leading to P2 and enhances the formation of closed complexes leading to P1. We have also to postulate that the activator complex favours the occupancy of the relevant promoter at malT by RNA polymerase. Hence, contrary to the action of the I repressor on AP,, , activation by the cyclic AMP-CRP complex acts very early in the kinetic pathway. Both at
lac and at gal it favours the formation of proper closed species able to isomerize quickly into the relevant active complexes. A possible mode of action of the cyclic AMP-CAP complex Two complementary modes of action of the activating complex are currently envisaged: protein-protein interactions could be decisive for the correct positioning of RNA polymerase in the closed-complex. This line of thought is supported by various indirect evidence: in solution cyclic AMP-CRP and RNA polymerase interact weakly and this interaction is reinforced when DNA is added (Blazy et al., 1980). Second, some effects of CRP on the location of RNA polymerase at gal promoters mutated in their CRP binding sites require that the two proteins assist each other in their mutual positioning (Spassky et al., 1984). Last, one mutant of CRP has been selected which still binds to the proper CRP binding site at lac but is unable to position RNA polymerase properly (N. Irwin, personal communication). The mutation has been directed to a site homologous to the one where it is assumed that the I repressor contacts RNA polymerase during the activation of the AP,, promoter (cf. Hawley & McClure, 1983; Hochschild et al., 1983). The complex between CRP and cyclic AMP does also bend the DNA. it does so at the three promoters studied here (cf. Kolb et al., 1983). A clear indication of this local bending has been provided by a variety of techniques (gel electrophoresis; cf. Wu & Crothers, 1984; Kolb et al., 1983; measurement of rotational relaxation times, Porschke et al., 1984; increase in the rate of formation of DNA minicircles; D. Kotlarz, unpublished work). There is no reason why these two phenomena should not participate together to the formation of the proper closed complex when the activator protein is present. Conclusion We propose that, during closed-complex formation, the cyclic AMP-CRP bends the promoter DNA and allows RNA polymerase to bind in its close vicinity by direct protein-protein interaction. This results in the formation of a nucleo-protein structure folding the DNA into a rather condensed configuration [A similar idea has been developed by Echols et al. (1984) for site specific recombination or initiation of DNA replication.] DNA would then be wrapped around part of the protein generating a negative superhelix. This local negative supercoiling will assist the formation of single-stranded regions at the rate limiting step, as suggested from the detailed study of the lac UV5 promoter. This model is mainly based on the work of my collaborators, A. Spassky, A. Kolb, S. Busby, M. Herbert, M. Amouyal and D. Kotlarz. It is a pleasure to thank all of them as well as W. McClure, P. Malan, K. Kirkegaard and J. Wang. The organizers acknowledge the support of I.C.I. plc towards the above lecture. Amouyal, M. & Buc, H. (1986) Biochem. Soc. Trans. 14, 272-273 Becker, M. M. & Wang, J. C. (1984) Nature (London) 309,682487 Blazy, B., Takahashi, M. & Baudras, A. (1980) Mol. Biol. Rep. 6, 3942 Buc, H. & McClure, W. R. (1985) Biochemisrry 24, 2712-2723 Chamberlin, M. J. (1974) Ann. Rev. Biochem. 43, 721-775 Chenchick, A., Beadealashvili, R. & Mirzabekow, A. (1981) FEBS L e r r . 128, 4 6 5 0
de Crombrugghe, B., Busby, S. & Buc, H. (1984) Science 224,831-837 Drew, H. R., Weeks, J. R. & Travers, A . A . (1985) EMBO J . 4, 1025-1 032
1986
614th MEETING, OXFORD
199
Echols, H., Dodson, M., Better, M., Roberts, J. D. & McMacken, R. (1984) Cold Spring Harbor Symp. Quant. Biol. 49, 727-733 Hawley, D. K. & McClure, W. R. (1983) Cell 32, 327-333 Hochschild, A,, Irwin, M. & Ptashne, M. (1983) Cell 32, 319-325 Kadesch, T. R., Rosenberg, S. & Chamberlin, M. J. (1982) J . Mol. Biol. 155, 1-29 Kirkegaard, K. & Wang, J. C. (1980) Cell 23, 721-729 Kirkegaard, K., Buc, H., Spassky, A. & Wang, J. C. (1983) Proc. Natl. Acad. Sci. U.S.A. 80, 2544-2548 Kolb, A. & Buc, H. (1982) Nucleic Acids Res. 10, 4 7 3 4 8 5 Kolb, A,, Spassky, A,, Chapon, C., Blazy, B. & Buc, H. (1983) Nucleic AcidF Res. 11, 7833-7852 Malan, T. P. & McClure, W. R. (1984) Cell 39, 173-180 Malan, T. P., Kolob, A,, Buc, H. & McClure, W. R. (1984) J . Mol. Biol. 180, 881-909 Maquat, L. & Reznikoff, W. (1978) J . Mol. Biol. 125, 4 6 7 4 9 0 McClure, W. R. (1985) Annu. Rev. Biochem. 54, in the press Porschke, D., Hillen, W. & Takahashi, M. (1984) EMBO J . 3, 2873-2878
Raibaud, 0. & Schwartz, M. (1984) Annu. Rev. Gen. 18, 173-206 Richmond, T. J., Finch, J. T. & Klug, A. (1986) Biochem. SOC.Trans. 14, OOO Roe, J. H., Burgess, R. R. & Record, T. R. (1984) J . Mol. Biol. 174 495-521 Rosenberg, S . , Kadesch, T. R. & Chamberlin, M. J. (1982) J . Mol. Biol. 155, 31-51 Siebenlist, U., Simpson, R. B. & Gilbert, W. (1980) Cell 20 269-281 Sigman, D., Spassky, A,, Rimsky, S & Buc, H. (1985) Biopolymers 24, 183-197 Spassky, A,, Busby, S. & Buc, H. (1984) EMBO J. 3, 1 4 3 Spassky, A,, Kirkegaard, K. & Buc, H. (1985) Biochemistry 24, 2723-2731 Stefano, J. E. & Gralla, J. D. (1982) Proc. Narl. Acad. Sci.U.S.A. 79, 1069-1 072 Wang, C. J. (1980) Trends Biochem. Sci. 5, 219-221 Wang, C. J., Jacobsen, J. H. & Saucier, J. M. (1977) Nucleic Acids Res. 4, 1221-1241 Wu, H. M. & Crothers, D. M. (1984) Nature (London) 308,509-513
DNA structure and promoter function ANDREW A. TRAVERS M.R.C. Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, U.K. Before RNA chain initiation, the product of the interaction of Escherichia coli RNA polymerase with promoter DNA is the separation of the DNA strands over a region 12 bp in extent in the region of the transcription startpoint. This implies that the net angular untwisting of the DNA duplex must amount to at least 410°, a value that is in good agreement with experimental determinations of the extent of topological unwinding during open-complex formation at several different promoters (Wang et al., 1977; Gamper & Hearst, 1982; Bertrand-Burggraf et al., 1984; Buc, 1986).
-
Untwisting How is the DNA converted from a fully base-paired duplex to a state in which slightly more than one turn is no longer base-paired and which properties of the promoter DNA itself are important in this process? Two such relevant properties are the ‘meltability’ (or more strictly, the probability of local untwisting) and the ‘bendability’. Untwisting is promoted by high temperature, certain solvents and negative supercoiling (Drew & Travers, 1984; Lee et al., 1981), and can be detected experimentally by the enhanced rate of cleavage of the unwound DNA by S1 nuclease (Drew, 1984; Drew et al., 1985) and by an increased probability of intercalation of the cleavage agent I , 10-penanthroline-copper ion (Sigman et al., 1985). One E. coli promoter whose activity is enhanced > 100 fold by negative supercoiling is that directing the expression of the tyrT transcription unit encoding a major tRNATy‘species (Lamond, 1985). When the DNA of this promoter is probed by S1 nuclease in the supercoiled form, we observe frequent cleavage sites that occur once every 30-50bp and are symmetrically placed on both strands (Drew et al., 1985). Each of these major cleavage sites is centred on a TA doublet which is the dinucleotide step with the lowest thermal stability (Gotoh & Tagashira, 1981). This result shows that untwisting occurs at preferred locations on the DNA duplex and that these locations are strongly sequence-dependent. In this experiment the site of greatest S1 cleavage is centred immediately 5‘ to the - 10 box, one of the two Abbreviation used: bp, base-pairs.
Vol. 14
highly conserved sequences known to be essential for efficient promoter function in E. coli (Hawley & McClure, 1983). It is therefore of interest to investigate whether any mutations in the - 10 box can directly affect the degree of untwisting and also the activity of the promoter. One such mutation is a T -+ A transversion converting the tyrT - 10 region from TATGAT to TATGAA (Berman & Landy, 1979). This mutation results in at least a 50-fold drop in the rate of initiation both in vivo (Berman & Landy, 1979) and in vitro (Travers et al., 1983). When unwinding of promoter DNA in the supercoiled state is assayed by S1 nuclease, the extent of cleavage in the - 10 region is reduced at least 10-fold by the mutation (Drew et al., 1985). The clear inference is that the mutant phenotype is, at least in part, a direct consequence of a change in the properties of the DNA in the absence of any interactions with RNA polymerase. Consistent with this conclusion is the observation that RNA polymerase bound to the mutant promoter always remains in the closed-complex configuration, i.e. in the state where DNA strand separation has not occurred (Travers et al., 1983). A further conclusion from this experiment is that the probability of unwinding at a particular site is dependent of the nature of the sequence flanking that site, since the position of the mutation is half a duplex turn distant from the position of maximal S1 cleavage. These experiments thus identify the proximal region of the - 10 box as the site in the tyrT promoter where the probability of local untwisting is greatest, and thus suggest that this is the region for the nucleation of untwisting leading to open-complex formation. Consistent with this model is the ubiquity of the TA doublet (often duplicated to TATA) in conserved regions of eubacterial, archaebacterial and eukaryotic promoters (Drew et al., 1985). This conclusion also indicates that since the melted region (- - 9 to + 3) in the open complex and the putative nucleation region (- - 9 to - 16) barely overlap, the untwisted region must migrate during open-complex formation (Fig. 1). Such a migration is again entirely consistent with the known geometry of polymerasepromoter complexes. In the initial closed complex the polymerase does not contact the DNA downstream of - 5 (Wang, 1982; Travers et al., 1983; Hofer et al., 1985). However in the closed complex the TA doublets on the transcribed strand in the - 10 region fall directly under the polymerase (Fig. 2). It is therefore reasonable
-
