Rho-dependent transcription termination

127

Biochimica et Biophysica Acta, 1048 (1990) 127-138 Elsevier

BBAEXP 92046

Review

Rho-dependent transcription termination John P. Richardson Department of Chemistry, Indiana University Bloomington, IN (U.S.A.) (Received 11 January 1990)

Key words: Transcription termination; RNA polymerase; Bacteriophage X; RNA dependent NTPase; RNA helicase

Contents I.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

127

II.

The structure of rho protein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

128

III.

Rho acts on the nascent RNA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

128

IV.

The termination stop points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

131

V.

Rho-mediated RNA release . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

131

VI.

R N A - r h o interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

133

VII.

Model for rho action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

135

VIII. Role of rho as a translational coupling factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

136

IX.

137

Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

137

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

137

I. Introduction

Individual genes or groups of genes are expressed by transcription of discrete segments of D N A molecules. Defined start and stop points for transcription are specified, respectively, by promoter and terminator sequences in the DNA. Although the relative level of expression of a gene is controlled primarily by the function of the promoter, the terminator limits the extent of the sequence transcribed. Thus, terminators serve a basic and essential function in the orderly

Abbreviation: NTPase, nucleoside triphosphate phosphohydrolase. Correspondence: J.P. Richardson, Department of Chemistry, Indiana University, Bloomington, IN 47405, U.S.A.

operation of a cell. In addition, they can also be important genetic regulatory elements that can modulate a graded level of expression of a set of genes controlled by a single promoter. In Escherichia coli, transcriptional termination is known to occur by two basically different mechanisms [1,2]. One mechanism allows a spontaneous release of the R N A transcript from R N A polymerase without the necessary involvement of a protein factor. The other mechanism involves the action of a release factor, a protein known as rho. The spontaneous release process occurs with a very circumscribed set of D N A sequences. In contrast, rho-dependent terminators occurs with a much greater variety of sequences. Thus, rho-dependent terminators can more readily be accomodated within other coding or regulatory signals. This review focuses

0167-4781/90/$03.50 © 1990 Elsevier Science Publishers B.V. (Biomedical Division)

128 on the mechanism of this diversely controlled form of transcription termination. Rho factor was discovered in 1969 by Jeffrey Roberts as a protein in E. coli that restricted transcription of bacteriophage ?t D N A in vitro to the earliest units of expression and that released the R N A products from the transcription complex [3], It was because of this release activity that it was assigned the greek letter rho (0) for identification as an E. coli R N A polymerase transcription factor. From the steady progress that has been made in the intervening two decades, the general outlines of its mechanism of action are evident. However, the mechanism is complicated and many aspects are only partially understood. Since rho plays an essential task in the process of gene expression, there is continued interest in the elucidation of the poorly understood aspects of its mechanism and it is the subject of intense research in several laboratories. Rho factor causes termination of transcription by first binding to a specific recognition site on a nascent R N A and then extending its interaction from that initial contact point toward the 3' end of the R N A with a process that is coupled to the hydrolysis of nucleoside triphosphates. The extension of the interaction essentially strips away the R N A polymerase and the D N A template to which the nascent R N A is attached in the transcription complex. Dissociation of the R N A from the R N A polymerase elongation site prevents further transcriptional elongation of the R N A chain. As with any biochemical process, the approach to understanding the various mechanisms involves studying the structural properties of the components and the partial reactions that are representative of the complete process. 11. The structure of rho protein Rho protein consists of a single subunit, the product of the rho gene, which readily associates to form multimeric structures up to a hexamer [4-6]. From the rhogene D N A sequence, it is known that rho protein consists of 419 amino acids [7], which is in good agreement with the estimates of its molecular weight from its gel electrophoretic migration [8]. Although it is not known tho have any modifications, it could possibly contain esterified acidic residues. This is suggested by the finding that isolated protein has measured p I value of 9 [9], yet the sequence predicts a total of 58 basic (Arg, Lys) residues and 59 acidic residues. Rho apparently functions as a hexamer. This inference is based primarily on the fact that it sediments as a hexamer when bound to poly(C) under conditions in which it is active [6]. However, in the absence of poly(C), it sediments as a dimer at low concentrations and as a tetramer at high concentrations [3,6]. Since cross-linking studies indicate that some fraction of free rho is hexameric, but no larger, the apparent mobility as

a tetramer is likely to be an average for a protein in which dimeric units are rapidly dissociating and reassociating from the hexameric form. In addition, electron micrographs of rho stained with uranyl acetate reveal a flat, ring-shaped structure composed of six globular units [10,11]. Each globular unit has a diameter of about 42 ,~ while the ring has a diameter of 119 A and a 30 hole in the center. Some specimens show a cleft on one side of the ring. Thus, instead of being a perfect ring, the six subunits may consist of one turn of a helix to form a structure that resembles a lock washer. Partial proteolysis and affinity-labeling techniques have indicated that a single subunit has three distinct structural domains [12]. The first domain, which extends from the amino terminus to a trypsin sensitive site at or near residue 130, contains a functional R N A binding site. This section from residue 20 to residue 90 has several sequence segments with similarity or identity with conserved elements of R N A binding proteins [13], while the section from residue 14 to residue 20 has a sequence that is similar to a sequence in the cytidine binding part of the regulatory subunit of aspartate transcarbamoylase. The second structural domain has a functional ATP binding site as well as most of the sequence segments with similarities and identities with conserved elements of nucleotide binding proteins [14]. The division between the second and third domains is a trypsin sensitive bond at residue 283 [15]. The third domain, which extends to the carboxyl terminus, does not become cross-linked by ultraviolet light-induced reaction with R N A or nucleotide ligands. However, it does contain some of the conserved sequence segments of N T P binding sites. In addition, mutants with altered functional properties have amino acid changes in this domain [16]. Although there is a distinct demarcation between the R N A and nucleotide binding domains, these two sections are not independent of each other; the ATP binding section is rapidly cleaved further by trypsin after a cut has been made in the bridging region and the cleavage at the bridging region is strongly suppressed when ATP is bound. A diagram representing the three domains is presented in Fig. 1. IlL Rho acts on the nascent R N A

The realization that an interaction with the nascent R N A is essential for termination of transcription by rho factor was inferred from the observations that the pure protein is an RNA-dependent nucleoside triphosphate phosphohydrolase (NTPase) [17] and termination was dependent upon the N T P hydrolysis reaction [18,19]. Early studies had already suggested that rho action in termination was not dependent upon some possible interaction with a specific D N A site [20], while the presence of RNase in reaction mixtures with rho present allowed R N A polymerase molecules to proceed

129 128 ~

tryosin

' ~ trypsin

Fig. 1. Diagram of the three distinct structural domains in a rho subunit. This is a summary of results presented in Ref. 12-16. Cyt. is used to represent a site that is specific for binding the cytidine portion of cytidylate residues (Richardson, LV. and Richardson, J.P., unpublished results), ill, f13, r4 and r5 are segments having sequences with similarity or identity with conserved elements of ATP-binding domains [14].

beyong rho-dependent terminators [21,22]. However, the most convincing demonstration of the role of binding to the nascent R N A came from the result that rho action was inhibited by D N A oligonucleotides that could bind to specific segments of the nascent R N A [23]. The oligonucleotides that inhibited rho action during transcription of the ~ cro D N A in vitro did not prevent its action at the l a c Z intragenic terminator during transcription of a l a c Z ' D N A fragment in vitro. Complexes between the isolated ~ cro R N A and the inhibitory oligonucleotides decreased the affinity of rho for R N A to a level that was characteristic of non-specific rhoR N A binding interaction [24]. Thus, a direct correlation could be made between a decrease in binding affinity and loss of termination function. Models could readily be proposed for involvement of an interaction between rho and the nascent R N A in termination [4,25]. Since each R N A molecule has a characteristic sequence and structure, the specificity of rho action in termination was readily accomodated by the nature and the stringency of the requirements for the interaction between rho and the RNA. At one extreme, rho could merely be binding to all R N A with no specificity. One variation of this type of model has rho start the interaction by binding to the 5' end of the nascent transcript [26]. At the other extreme, the binding could be to a very specific sequence. The evidence supports an intermediate between the extremes. There appears to be a requirement for a type of structure in the R N A with very limited constraints on the sequence. However, the stringency of the structural requirements is sufficient to account for a large part of the specificity of rho action.

A major reason for expecting that the structure of the R N A is more important for rho action than a particular R N A sequence came from a set of studies of the R N A requirements for activation of rho-ATPase. Purified rho protein catalyzes the hydrolysis of A T P as well as other nucleoside triphosphates when it can interact with certain R N A molecules [17,27]. Isolated R N A molecules transcribed from D N A molecules with rho-dependent terminators activate hydrolysis with turnover rate of as high as 48 ATP molecules per s per rho hexamer [24]. However, even higher rates are attained with synthetic R N A molecules such as poly(C) and copolymers of U or A with C, as cofactors [27]. These synthetic R N A molecules have no base-pairing. They also contain C residues. Other polymers with unpaired bases, such as poly(U) or poly(A) have very little cofactor activity. Thus, C residues are important for activation, but they need not be frequent because a U,C random copolymer with as few as 1 C residue for every 20 U residues is nearly as active as poly(C). The critical role of unpaired residues is demonstrated by the fact that the poly(C)poly(I) double-strand structure is inactive as are the highly structured Escherichia coil r R N A molecules. Three methods have been used to identify possible sequences and structures in the R N A transcripts that could be used by rho to make the initial binding interactions that can lead to transcription termination. The first involved analysis of the sequences of a collection of rho-terminated transcripts [28]. Since no elements of sequence regularity could be identified in this study, the sequences were also subjected to a program that was designed to predict the most stable combination of R N A base pairings. Each of the R N A molecules contained unstructured (non base-paired) regions of about 70-80 nucleotide residues in length within a few hundred residues of their 3' end. This analysis suggested that structure rather than sequence was the most important element. However, the significance and generality of this structural characteristic was not rigorously established because the analysis was not extended to a similar number of randomly chosen R N A transcripts that lack rho interaction sites and has not yet been tested for its ability to predict which transcripts should contain such a site. The second approach involved testing the effects of altering the upstream sequence segments in a gene containing a rho-dependent terminator. The most complete analyses by this approach have focused on ~ tR 1 [29,30], trpt' [31] and tyrTt [32]. Altered forms of the D N A templates were prepared in which segments of D N A were either deleted or replaced by segments of unrelated sequences. With MR1, two distinct 20-base pair sequence blocks, called rutA and rutB were shown to be particularly important for transcription termination with rho in vitro. The first block, rutA, started about 65 base pairs upstream from the first transcrip-

130

box B

A

A

box A

.C-..

Co

I I I I I

DNA

Fig. 2. Diagram of the nascent c r o RNA with RNA polymerase paused at site II of ktR a. This shows the close proximity of the c r o RNA rut segments to the RNA polymerase at a major termination stop point. The C residues in the two rut segments are circled to emphasize their possible importance. Although most are protected from cleavage by action of RNase CL3 when rho is bound [33], it is not known yet which are the most important for strong rho action at tR 1. The position of the single G residue in the two rut segments is also identified by a box.

tion stop point in tRy, while the second, rutB, ended about 10 to 15 base pairs upstream. The two blocks are separated by a 15 base pair segment that is distinctly less important than the flanking segments for rho action. Studies on the structure of the R N A [31] confirmed the expectation that the segments of the transcript from the two r u t blocks contain primarily unpaired nucleotide residues, while the segments between them, the r u t sequences, consist of a 5 base pair stem with a five residue loop. Fig. 2 shows a representation of the nascent ?~ c r o R N A attached to R N A polymerase paired at site II of tR 1 with the positions of the rutA and rutB segments delineated. The third method involved testing for blockage of rho function by oligonucleotides that bind to various segments of the nascent R N A [23]. The results obtained from using this approach showed that with ?~tRl, only oligonucleotides complementary to substantial part of the segment of the nascent R N A encoded by the rutA

and rutB sequences were effective in blocking rho action. Other oligonucleotides that readily form hybrid helical complexes with further upstream segments of the RNA, including the very 5' end segment [34], have no effect whatever on rho action. These results not only confirmed the importance of the rut segments, they were also consistent with the expectaton that rho interacts with the R N A at the rut segments and were inconsistent with models [26] in which a rho molecule initiates its interaction at the 5' end and translocates processively toward the 3' end. Although the position and function of the rut segments have been distinct for tR 1, the results represent a single case which may not be typical. Similar deletion analysis with trpt' did not reveal rut regions as distinct as those for XtR~ [31]. Thus, some nascent R N A molecules may have multiple, partially redundant, disperse segments that serve as the attachment-recognition elements of rho. The rut segment of tR~ immediately

131 preceeds the termination stop point. This situation is not necessarily a characteristic of the positioning of predicted rut elements for other terminators. However, without definitive location of rut sequences, it is premature to conclude that the relative position of rut site and termination stop points can vary considerably. A critical test of this relationship would be to test the effects of sequences inserted between rutB and the tR1 stop points. These experiments are in progress.

IV. The termination stop points The mechanisms that cause termination of transcription do not give rise to a homogeneous set of RNA molecules with identical 3' ends. Actual stop points for a terminator can be spread over several nucleotides [1,2]. Although this dispersion is small - involving only a few nucleotides - for a rho-independent terminator, it can range over 50 base pairs of DNA template for rho-dependent terminators [35-37]. Hence, with the requirement for an upstream rut region, a rho-dependent terminator comprises a minimum in some cases of 100 base pairs of D N A and could possibly range in size up to several hundred base pairs. Even with the most efficient of termination sites such as ~,tR 1, the stop points range over about 30 base pairs with stronger preferences at some points than others. The distribution of the preferences within that range follows closely the relative distribution of polymerase occupancy of the different sites during the transcription elongation process [38,39]. The stepwise elongation of RNA chains by R N A polymerase action does not occur with a constant time for each step [2]. Instead, the addition of some nucleotides occurs less rapidly than the addition of other nucleotides. Consequently, the enzyme 'pauses' briefly at some point along the DNA template. With several terminator regions, the transcription stop points correlate very well with the pause points. This observation is consistent with a model in which rho acts as a factor that can mediate release of RNA chains from transcription complexes. With this model, the frequency of utilization of a particular stop point would reflect the probability that RNA polymerase is at that point after rho has been able to bind to and act on the nascent RNA. Hence, a major determining feature of the exact position of the termination stop points is the way the DNA sequence influences the step time pauses. For many years, this was believed to be influenced primarily by the secondary structure characteristics of the segment of the RNA at the 3' end. There are many examples of pause sites that immediately follow segments of RNA that readily form base pair stem structures [2,40]. However, many cases are known where the pausing does not follow such structures [41,42]. The sequences and structural rules that govern elongation step time are not yet clear, but

they do appear to involve the D N A segment just preceding the nucleotides serving as the template [41]. Current work in progress in the laboratories of Chamberlin and Landick is focused on elucidating what the features are. Although the correlation between pause site and termination stop points is good within certain regions, it is not perfect [43]. One aspect of the transcription process that could clearly affect stop point preference is the stability of the ternary transcription complex. The fact that RNA polymerase terminates transcription by itself at certain sequences - at rho-independent terminators - indicates that there are some sequences where the ternary transcription complex is so unstable that release of RNA occurs spontaneously. Hence, there are likely to be other regions where the stability is too high for spontaneous release, but would provide less of a barrier for rho-mediated release than at other regions. Some of these distinctions are likely to be responsible for part of the selectivity that occurs in rho-mediated release of RNA from isolated transcription complexes. Recently, Arndt and Chamberlin (J. Mol. Biol., in press) have studied the intrinsic dissociation rate of RNA from transcription complexes paused at various regions on T7 templates and have found that the less stable complexes are invariably at positions where the 3' end of the nascent transcript can form a short stem structure (personal communication). They propose that the formation of RNA stems, which were once thought to be responsible primarily for pausing, may be more critical for facilitating release of a transcript.

V. Rho-mediated RNA release The activity of rho as an RNA release factor has been studied with two types of assay. One assay involves isolated transcription complexes which consist of DNA, RNA polymerase and nascent RNA with the RNA polymerase arrested at a variety of points on the DNA template [44,45]. The other measures dissociation of RNA molecules attached to single-stranded DNA through mutually complementary segments and does not involve RNA polymerase [4]. Since the dissociation of the RNA from the DNA in this second type of assay involves unwinding an R N A - D N A helix of 20 or more base pairs, it measures the ability of rho to function as an R N A - D N A helicase. The two types of assay have been helpful in establishing some important characteristics of the rho-mediated reaction that are directly related to its termination function. With the isolated transcription complexes, rho can mediate dissociation of nearly all transcripts from the complexes as long as the transcripts are larger than a certain size [44]. This minimal size presumably reflects a requirement for having enough binding components in the RNA for rho to make a sufficiently strong interac-

132 tion to initiate the release reaction. An essential requirement for release is the presence of a hydrolyzable nucleoside triphosphate [44,45]; analogs that are not substrates for rho NTPase are also inactive for release. The conditions that are optimal for release correspond well with conditions for N T P hydrolysis with isolated RNA, but are different from the conditions used for transcription. With ionic conditions used for transcription, R N A release from isolated complexes is more selective and the distribution of released transcripts parallels more closely the distribution of transcripts terminated by rho [45-48]. The key to the difference in selectivity is the concentration of Mg 2+ ions. Optimal release conditions were with 1 m M MgC12 [44]. However, RNA polymerase requires concentrations of 4 m M MgC12 or higher for rapid transcription [49] and the more selective release was found with 4 m M MgC12. The effects of MgC12 on ATP hydrolysis with isolated T7 R N A as a cofactor are very striking: the rate with 4 mM MgCI 2 is one-tenth the rate with 1 m M MgC12 [49]. If these differences in rate reflect in some way the ability of rho to act on the RNA, the loss of selectivity with the concentration of Mg 2+ ions that is optimal for ATPase and release activities could be a consequence of an improved ability of rho to dissociate the more tightly bound transcripts. Perhaps, therefore, a key element in the selectivity of rho action in transcription termination is the balance between the strength of rho's action on R N A and the strength of the attachment of R N A in the transcription complex. Experiments designed to monitor the release of D N A as well as R N A from R N A polymerase showed that the dissociation of D N A mediated by rho action closely paralleled the dissociation of the R N A [47]. Thus rho factor causes or allows separation of R N A polymerase from the template at or very near the time is separates the nascent R N A from the transcription complex. Further, the kinetics of release remained essentially coincident even using isolated transcription complexes that had been depleted of sigma factor [47]. Since sigma is known to enhance release of R N A polymerase from D N A at a rho-independent terminator [50], and to facilitate release of RNA polymerase bound to nonpromoter D N A sequences [51], it was expected to affect the release of polymerase from D N A at a rho-dependent terminus as well. The observation that it does not suggests that rho factor supplies the role of sigma by facilitating release of R N A polymerase core from the termination site D N A sequence. However, there is some concern that the D N A fragments used to trap the released R N A polymerase in these experiments may act to facilitate release of R N A polymerase after rho has pulled the R N A away by causing some sort of template displacement reaction that could be initiated at an end of a D N A molecule. Thus, the question of whether sigma is involved in a step of rho-dependent termina-

tion should be re-examined with the kinetic approach employed by Arndt and Chamberlin [50] to study release at a rho-independent terminator. The helicase assays have established some very critical features of rho action. First, the fact that rho can dissociate R N A from a D N A bound by a hybrid helix in a reaction that is coupled to ATP hydrolysis indicates that the release is the result of an active, physical process of stripping complexes attached at the 3' end of an R N A rather than the result of some conformational transition that the interaction of the rho-nascent R N A complex induces in the R N A polymerase molecules [46]. In other words, rho is not merely pushing a 'release button' on R N A polymerase. The possibility of the involvement of a specific rho-RNA polymerase interaction was first raised by early experiments which indicated that rho can bind to R N A polymerase [52]. More recently, it has been shown that rho can bind to NusA protein-RNA polymerase core complexes, but not very tightly to core R N A polymerase by itself [50]. However, since NusA protein is not required for the rho action, the significance of that observation is not understood. A specific rho-RNA polymerase interaction was also inferred from some genetic analyses. When mutants are selected that have regained some of the lost termination function that is the characteristic of a mutated rho factor, suppressors are found that have mutations in the gene for the/3-subunit of R N A polymerase [54]. However, it is possible that these rho-mutant suppressor variants of R N A polymerase just have slower rates of R N A elongation or weaker R N A binding characteristics, which would compensate for the slower or weaker action of defective. Indeed recent results by Jin and Gross (personal communication) have suggested that one of the suppressor variants of R N A polymerase does have a slower rate of chain growth. Since the mechanism for suppression can be explained in terms of relative rates of action rather than specific rho-polymerase contacts, this is no longer a strong reason to consider rho anything other than a non-specific release factor. Hence, this result along with the results of the helicase activity makes the involvement of specific rhoR N A polymerase interactions seem unlikely. The helicase assays also established the polarity of rho action along the R N A [46]; D N A attached by a hybrid helix at the 3' end of the target R N A were released while those at the 5' end were not. This suggests that rho acts with a net 5' to 3' direction along the R N A and is the major evidence that is consistent with an R N A tracking mechanism that is coupled to ATP hydrolysis. An extremely useful feature of the helicase assay is that it simplifies the task of testing the ability of rho to act on various R N A - D N A complexes in a reaction that is closely related to the natural termination event. It can be used, for instance, to test systematically the effect of

133 secondary structural elements in the R N A on release activity. It also may provide the key test of the tracking mechanism model. VI. R N A - r h o interactions

Since isolated R N A molecules are sufficient for activating A T P hydrolysis catalyzed by rho, some of the basic interactions that cause rho to release R N A have been deduced from studies of r h o - R N A interactions. This approach became feasible with the development of transcriptional expressions systems that could be used to synthesize mg quantities of specific R N A molecules. However, some properties were revealed from studies with simple sequence R N A molecules such as poly(C), poly(U) or poly(A). Rho binds very rapidly to isolated R N A transcripts [55]. Actually, the binding reaction is so fast that it has not been determined by the conventional binding assays that have been utilized so far. The complexes with ~, cro R N A form in less than 10 s, which, with the concentration of rho and R N A used, indicate that the bimolecular association rate constant (ka) is greater than 7-107 M - 1 . s -1. This much, at least, is consistent with a critical involvement of the rho-nascent R N A interaction in transcription termination. A result that was less expected, however, is that the rate of dissociation is also too fast to measure by the standard filter binding assay; when complexes were diluted to a concentration below which rho was unable to rebind, rho was released from the R N A before the first sample could be filtered 10 s later (57). The equilibrium association constant for the r h o - c r o interaction ( K a ) is 8.0"108 M -1 at 3 7 ° C [24,55]. Thus, this relatively tight interaction is a consequence of fast on and off rates. These measurements represent binding of rho to R N A in the absence of ATP. However, in the presence of ATP, no change could be detected in the overall binding affinity [55]. This result is surprising because actions of rho on the R N A that are coupled to ATP hydrolysis would likely perturb some aspect of the association or dissociation process. The results, or at least their interpretation, may be subject to an experimental flaw; these binding measurements were performed with a very low concentration of R N A and stoichiometric excess of rho, i.e., binding was driven by the rho concentration, and although ATP may have been present, it may not have been hydrolyzed at the rate expected for normal turnover be unhindered enzyme. At high concentrations of RNA, stoichiometric excess of rho is known to inhibit ATP hydrolysis by the complexes (Richardson, L.V. and Richardson, J.P., unpublished data), a result which suggests that the binding of a second rho molecule to the R N A could interfere with the action of rho bound initially. Thus, the effect of A T P hydrolysis on binding equilibrium

needs to be remeasured under experimental conditions in which ATP hydrolysis can actually be shown to be the unhindered rate. From some experiments performed with poly(C), it is known that coupled A T P hydrolysis does affect r h o - R N A interaction [56]. Rho binds very tightly to poly(C), so tightly, in fact, that a reliable association constant has not yet been measured. It is at least as high as 3 • 1 0 9 M - 1 and possibly higher [56,57]. Under most binding conditions, the yields of complexes reflect the stoichiometry of binding rather than the affinity with poly(C). This stoichiometry is not altered by ATP. The binding to poly(C) is strong enough that dissociation rates can be measured by a membrane filtration technique. The first-order rate constant for dissociation of rho-poly(C) at 37 ° C in 0.04 M Tris-HC1 (pH. 8.0), 0.25 m M KC1 and 10 m M MgC12 is 2.6-10 -4 s i in the absence of ATP and 5.7. ] 0 - 4 S - 1 in the presence of ATP [56]. Thus, some step coupled to A T P hydrolysis appears to enhance by a factor of two the release of rho from a poly(C) molecule. The slow rate of release of rho from poly(C) has made it possible to determine the size and extent of R N A that can be protected from degradation with pancreatic ribonuclease. The extent is 60 cytidylate residues per rho hexamer in fragments that are about 60-80 residues in length [11,56]. This suggests that each hexamer has a single, well protected site that extends across all six subunits. This extent of protection is very similar to the site size of rho occupancy measured by titration of the amount of rho that can saturate the poly(C) in the absence of A T P [56,57]. The extent and size of poly(C) protected from digestion with pancreatic RNase did not change in the presence of ATP. In contrast, the amount of rho needed to saturate poly(C) decreased by a factor of two when A T P was added [56]. Thus, fewer rho molecules can fit on a poly(C) when ATP hydrolysis is occurring even though the extent of protection remains the same. The result suggests that the rho-poly(C) interaction changes with A T P hydrolysis. One type of mechanism that could account for the change in interaction involves binding of poly(C) at two kinds of sites in rho. One is a large site - large enough to bind up to 60 contiguous C residues - that would extend across all 6 subunits of the hexamer. Poly(C) would bind to the site in the absence of A T P and would remain bound to it even when A T P is being hydrolyzed. The second site would be smaller and make transient interactions coupled to the ATP hydrolysis cycle. These transient interactions either wind the poly(C) loosely around the outside or causes rho to track along the poly(C). By either mechanism, fewer rho molecules could occupy space at poly(C) during A T P hydrolyis without involving a significant (or detectable) increase in the extent of protection from R N a s e action. Because the six

134 subunits of rho are identical, it is reasonable to propose that there may be one secondary site per subunit. However, hexameric rho binds only three molecules of ATP [58], thus by extension, it may have only three secondary sites, one for every pair of subunits. Further evidence in support of the two site model came from the finding that ATP hydrolysis can be activated by a mixture of single-stranded D N A and oligoribonucleotides [59]. Rho can bind to singlestranded D N A in the absence of ATP and that binding is competitive with R N A [56]. However, D N A alone does not activate ATP hydrolysis. Thus an interaction with R N A is needed to activate ATP hydrolysis. Since D N A can bind to the primary site, the secondary sites could be the ones specific for R N A molecules. According to the two site model, interactions with the secondary site should be coupled to the ATP hydrolysis cycle. The finding that the g m for the oligoribonucleotides in activation of ATP hydrolysis with poly(dC) depends upon the ATP concentration is fully consistent with this expected coupling [59]. In spite of this encouraging evidence, all attempts to detect the direct binding of oligoribonucleotides to rho in the presence of poly(dC) have been unsuccessful [12]. These attempts have included the use of ATP, non-hydrolyzable ATP analogs, and ADP. Conceivably, binding to this site occurs only when rho is in some intermediate state of the ATP hydrolysis cycle. Attempts are now being made to find ATP analogs that might lock rho in the conformation of the intermediate state. The binding of rho of k c r o R N A has been analyzed with respect to the relative role of the rut regions and the contribution of ionic and non-ionic bonds [24]. The change in standard free energy (AG O' ) of the binding reaction of rho with h c r o R N A is 12.6 kcal/mol. Of that total, about one-third is due to formation of ionic bonds and two-thirds to non-ionic interactions (hydrogen bonds and Van der Waals bonds). Rho binds with 6-fold lower affinity to a variant of c r o R N A lacking all of one rut segment and most of the other rut segment than to normal c r o RNA. This difference amounts to 1.1 kcal lower (less negative) binding enery, which is due almost completely to a loss of non-ionic binding energy, and correlates with a nearly total loss of termination at tR 1 during transcription of a D N A template encoding the synthesis of this same variant RNA. A similar decrease in overall binding energy is also achieved with the normal c r o R N A itself by increasing the KC1 concentration from 0.05 M to 0.15 M, although in this case, the decrease is due to a weaker ionic interaction caused by counterion competition. Again, this decrease in binding energy of about 1 . 2 + 0 . 2 k c a l / m o l correlates with a nearly total loss of termination activity. However, these changes in sequences or conditions also affect other aspects of the functional interaction besides binding or rho to the R N A . This is

evident from the finding that the Vmax for ATP hydrolysis by rho at R N A saturation is 10-fold lower with either the variant c r o R N A in 0.05 M KCI or normal c r o R N A in 0.15 M KC1 than with normal c r o R N A in 0.05 M KC1 [24]. Since defects in mere binding affinity would presumably be overcome by higher concentrations of the RNA, the alterations are affecting other steps that are closely coupled to ATP hydrolysis. If secondary interactions of rho with the R N A are truly involved, these interactions could be compromised by the high KCi concentration or by the lack of the R N A sequences in the variant RNA. Nonetheless, the binding studies indicate clearly that the presence of the rut sequences has an important effect on the primary binding. Direct evidence for a physical contact of rho with the rut segments of c r o R N A have come from the use of chemical and enzymatic probes of the accessibility of specific R N A residues [33]. Many of the C residues and the one G residue in the rut region of ~ c r o R N A are much less readily cleaved by ribonucleases in the presence of rho than in its absence, whereas cleavage at other accessible residues outside of the rut region remain unchanged. The ability to detect a physical contact between rho and the rut region by this 'footprint' type of analysis allowed a test of two alternative mechanisms for the interaction between rho and R N A that are coupled to ATP hydrolysis. With one possible mechanism, ATP hydrolysis is coupled with a translocation of rho from the initial attachment site in the rut region toward the 3' end of the R N A [25]. With the other, hydrolysis is coupled with the formation of additional contacts near the 3' end while maintaining an attachment at the rut region [56]. The results showed that a G residue at the 5' end of the 5' rut segment of X c r o R N A that becomes blocked from action of T1 RNase with rho present remains inaccessible when ATP is being hydrolyzed by the r h o - c r o R N A complex [33]. This result suggests that rho maintains an attachment with the rut region while it is making interactions with RNA that are coupled to ATP hydrolysis. However, since very little is known about the dynamics of the coupled A T P hydrolysis reaction, the time taken for motion and release could be short compared to the time spent at the initial attachment site. Hence, this experiment does not eliminate certain variants of the translocation model. An attractive feature of the mechanism in which rho maintains a contact with a rut region is that the special binding energy associated with the contacts at the rut region can be used to enhance the more transient interactions at other points along the RNA. The binding studies indicated that the rut region contributes some important non-ionic interactions to the overall binding energy [24]. These non-ionic bonds are likely to involve interactions with the C residues. This interpretation is suggested from two types of experiments. One is that

135 the binding of rho of poly(C) is insensitive to salt concentration [56]. The other is that the binding of rho to A cro R N A can be inhibited by cytidine, a non-ionic molecule (Richardson, L.V. and Richardson, J.P., unpublished results). Studies of the relative ability of various cytidine analogs to inhibit the r h o - c r o R N A interaction have suggested that rho contains a cleft that makes specific H-bond interactions with the C 2 keto oxygen and C4 amino groups of cytidine and makes a close Van der Waals contact with the edge of the ribose ring in the C~ carbon (Richardson, L.V. and Richardson, J.P., unpublished results). Because rho contains a small sequence in its R N A binding domain that has several identities with a sequence segment of that former part of the CTP binding site of aspartate transcarbamylase [12], it will be of interest to determine whether residues in the sequence also form part of the C-specific site of rho factor.

~

rho- 3 ATP

7

nATP + nil20

~

nADP + nPi

VII. Model for rho action .

The elements of a model for rho action are diagrammed in Fig. 3. The first panel represents an R N A polymerase molecule in the process of making an R N A transcript. The enzyme has reached a point at which segments of the RNA, the rut segments, are available for a strong binding interaction with the primary binding site of rho. This site extends across all six subunits and has three clefts, one per dimer, each capable of binding an unpaired cytidylate residue. In this example, two rut segments bind into two thirds of the total primary site. The extent of occupancy will vary from one R N A to the next and would be a determining factor of the effectiveness of rho action. Since rho binds to NTPs (shown as ATP because it has the highest affinity) in the absence of R N A [12,58], a complex of rho with 3 ATP molecules binds to RNA. The presence of R N A in the primary site and ATP molecules in their sites initiates a conformational change that extends the contact of R N A to sites that are operationally termed secondary R N A binding site and that simultaneously brings the residues that catalyzes hydrolysis into alignment to allow attack of solvent O H - on the "f phosphorous atom of an ATP molecule in a coupled site [61]. The extension of the R N A contact displaces base pair bonds of the R N A with other segments of the R N A and with the segment of the template D N A strand. The model is necessarily vague about the number, position and timing of the secondary contact and, consequently, the number of ATP molecules hydrolyzed. The diagram shows two secondary site contacts for illustrative purposes and gives the stoichiometry of ATP hydrolysis as an indeterminant value, n. Also, since it is not known whether rho actually contacts the hybrid helix between the 3' end of the R N A and the D N A template or whether it merely contacts the surface

~

rho-RNA + RNA polymerase-DNA rho + RNA

RNA polymerase + DNA

Fig. 3. A model for rho action. This diagram shows a prog#'essionof steps that starts with RNA polymerase at or near a pause site-

termination stop point with a nascent RNA containing two rut segments of about 20 nucleotides each (patterned after the rut segments of A cro RNA). Rho, which can bind ATP (or any NTP) in the absence of RNA, binds to the RNA-RNA polymerasecomplex as a 3 ATP-rho complex, with major contacts between segments of the large, primary site in rho and the rut segments on the nascent RNA. In steps that are coupled with the hydrolysisof the bound ATP molecules, rho is then shown making two additional, secondary site contacts with segments of the nascent RNA toward its 3' end. These secondary contacts replace RNA-RNA base pair interactions that were in an intervening stem as well as RNA-DNA base pairs in the transcription complex. Possibly, a Newtonian force of the rho contacting the RNA polymerase coupled with the binding contacts of the nascent RNA with rho helps to dislodge the transcript from the DNA and the RNA polymerase.

of the R N A polymerase at the exit of the R N A product site, this latter case is diagrammed because the hybrid helix between the nascent transcript and the template D N A does appear to be largely inaccessible to penetration by other proteins, such as nucleases [62]. The disruption of the helix between the nascent R N A and the D N A template weakens the overall interaction of

136 the R N A with R N A polymerase and since R N A does not stay tightly bound to free rho [55], it will be released from the entire complex. From the evidence that sigma factor is not essential for rapid release of D N A from the transcription complex at a rho-dependent terminator, the contacts between rho and R N A polymerase may also weaken the binding contact between core R N A polymerase and the D N A at the termination stop point. The actual position of the termination stop point depends on the dynamics of binding of rho to the rut segment, the rate with which the contacts can be extended toward the 3' end of the RNA, the rate with which R N A polymerase continues to add nucleotides and the relative affinity constants. Thus rho could start its interaction with the nascent R N A before R N A polymerase has reached the pause sites that become the release points. The actual position of a stop point depends upon the strength of interactions that govern the chain elongation by R N A polymerase, the structure of the R N A transcript and the strength attachment of the nascent R N A to the R N A product site of R N A polymerase and the template D N A strand. Although there is no strong evidence that rho factor affects directly the R N A chain elongation process, its ability to unwind R N A - R N A helix structures could affect the structure of the nascent R N A in the R N A product site and thus affect the chain elongation indirectly. But clearly the main involvement of rho in transcription termination as a release factor and this is governed primarily by its ability to extend contact toward the 3' end of the nascent RNA. VIII. Role of rho as a translational coupling factor

One interesting question concerning the mechanism of rho action is why does E. coli, and presumably other bacteria, have such terminators when it is possible to terminate transcription by a simpler, intrinsic mechanism. One possible answer is that a factor, such as rho, that acts on the nascent R N A could serve to distinguish between translated and nontranslated portions of the R N A and thereby couple continued transcription to a functional utilization of the nascent R N A . Evidence for this role is suggested by the finding that many genes contain latent, rho-dependent transcriptional terminators [63]. These terminators give rise to a genetic phenomenon called polarity. Certain mutations in genes that yield incomplete products due to premature translation termination often decrease the expression of downstream genes in the same operon. This decrease in expression is due to termination of transcription at latent terminators within the gene with the primary lesion. These terminators become active because premature translational termination allows the R N A to be synthesized without ribosomes present throughout the length of the transcript. Presumably, the

presence of a ribosome translating the nascent m R N A as it emerges from the R N A polymerase either blocks rho from binding to the nascent R N A or blocks access or rho to R N A polymerase along the RNA. However, if the ribosome is absent and the R N A contains a rut site, rho will be able to bind and act [4]. Since polar mutations have been found in m a n y E. coli genes, particularly those that are the first to be translated from a polycistonic m R N A [63], latent transcriptional terminators appear to be a general feature of genetic organization that may serve some important physiological role. Bacteria, such as E. coli, are well adapted for sharp changes in the availability of nutrients. Starvation for an amino acid will eventually lead to induction of the enzymes that catalyze synthesis of that amino acid, but during the induction period the cell needs to conserve energy resources and not make R N A that will be translated inefficiently. One consequence of a temporary imbalance in which the rate of translation drops but the rate of transcription elongation does not is that the ribosomes will no longer be present on the segment of nascent R N A that emerges from R N A polymerase. In this case, latent rut sites would be available for binding with rho, which could therefore act in turn to terminate transcription. Evidence for a starvation-induced termination of transcription at latent rho-dependent terminators has been obtained in studies with lacZ gene expression [43]. Starvation for seryl-tRNA induced by addition of serine hydroxamate, a competitive inhibitor of seryl-tRNA synthetase, caused a 2.5-fold reduction in the relative amount of R N A from a segment just downstream from the set of terminators in the first 500 bp of the lacZ gene. This treatment also caused an increase in the fraction of transcripts with 3' ends at or near nucleotide 220 of lacZ m R N A , which is one of the major end points of the latent lacZ intragenic terminators. Since neither of these responses occurred in an isogenic strain with a defective rho factor, they are likely to be due to actions of rho factor on the latent terminators. This evidence is consistent with the proposal concerning the functional relevance of latent intragenic terminators. Although the effects of starvation on continued transcription of lacZ were evident there are also obvious effects on the overall level of expression that arise primarily because of the severe catabolite repression that sets in when cells are starved for amino acids [64]. Rho's ability to abort continued transcription of R N A molecules that are not translated well is just one of several mechanism used to protect the cell from the effects of starvation. In the course of testing segments of the lacZ gene for the presence of latent terminators, several were found in the anti-sense orientation of the gene (unpublished data). Because of the diverse and non-specific sequence requirement for a rho-dependent terminator, this observa-

137 tion is not surprising. The presence of such terminators may serve a very useful role: they would function to abort continued elongation of R N A polymerase molecules that initiated transcription by mistake at a cryptic promoter. Although it is not known how frequently mistakes are made in the recognition of promoters, the sequences that can be used to initiate transcription are diverse and minor mutational changes are known to convert a non-promoter sequence to an active promoter [65]. One clear consequence of the function of a termination factor that acts by binding to an unencumbered nascent transcript is that an R N A chain initiated from the antisense strand of a gene would produce transcripts that would not be continually translated would be a target for such a factor. Terminators at the ends of genes or gene groups can either be rho-dependent or rho-independent. AtR 1 and trpt' are clear examples of rho-dependent terminators at the end of a gene or a group of genes. These terminators consist of 40 or more rut site sequences plus the termination stop point regions, which can extend over some 40 or more base pairs. In contrast, the sequence of a rho-independent terminator, which consists of a G-C rich segment of about 20 to 25 bp containing an interrupted dyad symmetry followed by a 7 bp stretch with mainly dA residues on the template strand, is not as extensive as a rho-dependent terminator but has a more rigid sequence requirement. Either type could have served as the exclusive type of terminator at the end of genes, but because of the special sequence requirements of the rho-independent terminators, they would be more difficult to adapt to a coding sequence. Hence, rho-dependent terminators are better suited to the task of acting to terminate transcription of R N A molecules that are not being translated properly or that are being made by mistake than are rho-independent terminators.

IX. Concluding remarks Rho belongs to a large class of proteins that couple hydrolysis of NTPs with a physical action on a nucleic acid. This class includes the D N A helicases that are involved in unwinding the D N A fork [66-68] and other steps of D N A replication [69] and the translational initiation factors elF4A and elF4F [70] that either help 40 S ribosomes to bind to eucaryotic m R N A or actually propel the subunit along the R N A during its scan for the first A U G sequence in the proper context [71,72]. However, with none of these proteins is the mechanism of action understood. Although some aspects of the rho mechanism will be unique for it as a transcription termination factor, other aspects, such as the ability to displace base pair contacts with protein-nucleotide residue contacts, will likely be similar for most members of this class of proteins. Working out the mechanism for rho factor therefore will not only reveal how termina-

tion occurs, but will also reveal secrets to key steps in replication and translation. Acknowledgements Research cited from my laboratory was supported by N1H Grant AI 10142. I thank Kelly Blackwell for preparing the manuscript.

References 1 Platt, T. (1986) Annu. Rev. Biochem. 55, 339-372. 2 Yager, T.D. and Von Hippel, P.H. (1987) in Escherchia coli and Salmonella Typhimurium, Cellular and Molecular Biology (Neidhardt, F.C., ed.), pp. 1241-1275, American Society for Microbiology, Washington, D.C. 3 Roberts, J.W. (1969) Nature 224, 1168-1174. 4 Richardson, J.P., Grimley, C. and Lowery, C. (1975) Proc. Natl. Acad. Sci. USA 72, 1725. 5 Ratner, D. (1976) in RNA Polymerase (Losick, R. and Chamberlin, M., eds.), pp. 645-655, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY. 6 Finger, L.R. and Richardson, J.P. (1982) J. Mol. Biol. 56, 203-219. 7 Pinkham, J.L. and Platt, T. (1983) Nucleic Acids Res. 11, 3531-3545. 8 Finger, L.R. and Richardson, J.P. (1981) Biochemistry 20, 1640-1645. 9 Blumenthal, R.M., Reeh, S. and Pederson, S. (1976) Proc. Natl. Acad. Sci. USA 73, 2285-2288. 10 Oda, T. and Takanami, M. (1973) J. Mol. Biol. 71, 799-802. 11 Bear, D.G., Hicks, P.S., Escudero, K.W., Andrews, C.L., McSwiggen, J.A. and Von Hippel, P.H. (1988) J. Mol. Biol. 199, 623-635. 12 Dolan, J.W., Marshall, N.F. and Richardson, J.P. (1990) 265, in press. 13 Query, C.C., Bentley, R.C. and Keene, J.D. (1989) Cell 57, 89-101. 14 Dombroski, A.J. and Platt, T. (1988) Proc. Natl. Acad. Sci. USA 85, 2538-2542. 15 Bear, D.G., Andrews, C.L., Singer, J.D., Morgan, W.D., Grant, R.A., Von Hippel, P.H. and Platt, T. (1985) Proc. Natl. Acad. Sci. USA 82, 1911-1915. 16 Mori, H., Imai, M. and Shigesada, K. (1989) J. Mol. Biol. 210, 39-50. 17 Lowery-Goldhammer, C. and Richardson, J.P. (1974) Proc. Natl. Acad. Sci. USA 71, 2003-2007. 18 Howard, B.H. and De Crombrugghe, B. (1976) J. Biol. Chem. 251, 2520-2524. 19 Galluppi, G.R., Lowery, C. and Richardson, J.P. (1976) in RNA Polymerase (Losick, R. and Chamberlin, M., eds.), pp. 657-665, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY. 20 Richardson, J.P. (1970) Cold Spring Harbor Symp. Quant. Biol. 35, 127-133. 21 Darlix, J.L. (1973) Eur. J. Biochem. 35,517-526. 22 Sharp, J.A. and Platt. T. (1984) J. Biol. Chem. 259, 2268-2273. 23 Chen, C.-Y.A., Galluppi, G.R. and Richardson, J.P. (1986) Cell 46, 1023-1028. 24 Faus, I. and Richardson, J.P. (1989) Biochemistry 28, 3510-3517. 25 Adhya, S. and Gottesman, M. (1978) Annu. Rev. Biochem. 47, 967-996. 26 Lewin, B. (1987) Genes, Third Edition, p. 260, John Wiley & Sons, New York. 27 Lowery, C. and Richardson, J.P. (1977) J. Biol. Chem. 252, 1381-1385. 28 Morgan, W.D., Bear, D.G., Litchman, B.L. and Von Hippel, P.H. (1985) Nucleic Acids Res. 13, 3739-3754. 29 Lau, L.F. and Roberts, J.W. (1985) J. Biol. Chem. 260, 574-584.

138 30 Chen, C.A. and Richardson, J.P. (1987) J. Biol. Chem 262, 11292-11299. 31 Galloway, J.L. and Platt, T. (1988) J. Biol. Chem. 263, 1761-1767. 32 Madden, K.A. and Landy, A. (1989) Gene 76, 271-280. 33 Fans, I. and Richardson, J.P. (1990) J. Mol. Biol., in press. 34 Fans, I. (1988) Ph. D. dissertation, Indiana University, Bloomington, IN. 35 Wu, A.M., Christie, G.E. and Platt, T. (1981) Proc. Natl. Acad. Sci. USA 78, 2913-2917. 36 Lau, L.F., Roberts, J.W. and Wu, R. (1982) Proc. Natl. Acad. Sci. USA 79, 6171-6175. 37 Morgan, W.D., Bear, D.G. and Von Hippel, P.H. (1983) J. Biol. Chem. 258, 9553-9564. 38 Lau, L.F., Roberts, J.W. and Wu, R. (1983) J. Biol. Chem. 258, 9391-9397. 39 Morgan, W.D., Bear, D.G. and Von Hippel, P.H. (1983b) J. Biol. Chem. 258, 9565-9574. 40 Landick, R. and Yanofsky, C. (1987) in Escherichia coil and Salmonella typhimurium, Cellular and Molecular Biology (Neidhardt, F.C., ed.), pp. 1276-1301, American Society for Microbiology, Washington, DC. 41 Levin, J.R. and Chamberlin, M.J. (1987) J. Mol. Biol. 196, 61-84. 42 LaFlamme, S.E., Kramer, F.R. and Mills, D.R. (1985) Nucleic Acids Res. 13, 8425-8440. 43 Ruteshouser, E.C. and Richardson, J.P. (1989) J. Mol. Biol. 208, 23-43. 44 Richardson, J.P. and Conaway, R. (1980) Biochemistry 19, 4293-4299. 45 Shigesada, K. and Wu, C.-W. (1980) Nucleic Acids Res. 8, 3355-3369. 46 Brennan, C.A., Dombroski, A.J. and Platt, T. (1987) Cell 48, 945-952. 47 Andrews, C. and Richardson, J.P. (1985) J. Biol. Chem. 260, 5826-5831. 48 Morgan, W.D., Bear, D.G. and von Hippel, P.H. (1984) J. Biol. Chem. 259, 8664-8671. 49 Richardson, J.P. and Macy, M.R. (1981) Biochemistry 20, 1133-1139. 50 Arndt, K.M. and Chamberlin, M.J. (1988) J. Mol. Biol. 202, 271-285. 51 Hinkle, D.C. and Chamberlin, M.J. (1972) J. Mol. Biol. 70, 157-185.

52 Darlix, J.L., Sentenac, A. and Fromageot, P. (1971) FEBS Lett. 13, 165-168. 53 Schmidt, M.C. and Chamberlin, M.J. (1984) J. Biol. Chem. 259, 15000-15002. 54 Das, A. Merril, C. and Adhya, S. (1978) Proc. Natl. Sci. USA 75, 4828-4832. 55 Ceruzzi M., Bektesh, S.L. and Richardson, J.P. (1985) J. Biol. Chem. 260, 9412-9418. 56 Galluppi, G. and Richardson, J.P. (1980) J. Mol. Biol. 138, 513-539. 57 McSwiggen, J.A., Bear, D.G. and yon Hippel, P.H. (1988) J. Mol. Biol. 199, 609-622. 58 Stitt, B.L. (1988) J. Biol. Chem. 263, 11130-11137. 59 Richardson, J.P. (1982) J. Biol. Chem. 257, 5760-5766. 60 Ref. deleted. 61 Stitt, B.L. and Webb, M.R. (1986) J. Biol. Chem. 261, 15906-15909. 62 Rohrer, H. and Zillig, W. (1977) Eur. J. Biochem. 79, 401-409. 63 Adhya, S., Gottesman, M., de Crombrugghe, B. and Court, D. (1976) in RNA Polymerase (Losick, R. and Chamberlin, M., eds.), pp. 719-730, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. 64 Magasanik, B. (1970) in Lactose Operon (Beckwith, J.R. and Zipser, D., eds.), pp. 189-219, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. 65 Hawley, D.K. and McClure, W.R. (1983) Nucleic Acids Res. 11, 2237-2255. 66 LeBowitz, J.H. and McMacken, R. (1986) J. Biol. Chem. 261, 4738-4748. 67 Baker, T.A., Sekimizu, K., Funnell, B.E. and Kornberg, A. (1986) Cell 45, 53-64. 68 Challberg, M.D. and Kelly T.J. (1989) Annu. Rev. Biochem. 58, 671-717. 69 Lasken, R.S. and Kornberg, A. (1988) J. Biol. Chem. 263, 5512-5518. 70 Ray, B.K., Lawson, T.G., Kramer, J.C., Cladaras, M.H., Grifo, J.A., Abramson, R.D., Merrick, W.C. and Thach, R.E. (1985) J. Biol. Chem. 260, 7651-7658. 71 Kozak, M. (1980) Cell 22, 459 467. 72 Kozak, M. (1989) J. Cell Biol. 108, 229-241.