Mechanism for a transcriptional activator that works at the isomerization step Simon L. Dove, Franklin W. Huang, and Ann Hochschild* Department of Microbiology and Molecular Genetics, Harvard Medical School, Boston, MA 02115 Edited by Mark Ptashne, Memorial Sloan–Kettering Cancer Center, New York, NY, and approved September 26, 2000 (received for review June 30, 2000) Transcriptional activators in prokaryotes have been shown to stimulate different steps in the initiation process including the initial binding of RNA polymerase (RNAP) to the promoter and a postbinding step known as the isomerization step. Evidence sug- gests that activators that affect initial binding can work by a cooperative binding mechanism by making energetically favorable contacts with RNAP, but the mechanism by which activators affect the isomerization step is unclear. A well-studied example of an activator that normally exerts its effect exclusively on the isomer- ization step is the bacteriophage l cI protein (lcI), which has been shown genetically to interact with the C-terminal region of the s70 subunit of RNAP. We show here that the interaction between lcI and s can stimulate transcription even when the relevant portion of s is transplanted to another subunit of RNAP. This activation depends on the ability of lcI to stabilize the binding of the transplanted s moiety to an ectopic 235 element. Based on these and previous findings, we discuss a simple model that explains how an activator’s ability to stabilize the binding of an RNAP subdomain to the DNA can account for its effect on either the initial binding of RNAP to a promoter or the isomerization step. Many transcriptional activators in prokaryotes bind to spe- cific sequences associated with the promoters they regu- late and affect the initiation process through direct contacts with RNA polymerase (RNAP; subunit structure, a2bb9s) (1–3). The process of transcription initiation in Escherichia coli can be described by a simplified two-step model (4). First, RNAP binds to fully duplex promoter DNA to form what is called the closed complex. Formation of this complex is reversible and is described by an equilibrium binding constant KB. For transcription to initiate, the closed complex must then isomerize to form the transcriptionally active open complex in which the DNA is locally melted to expose the transcription start site. This isomerization step is usually irreversible and is described by a forward rate constant kf. The cAMP receptor protein (CRP) is a well- characterized example of an activator that can exert its effect exclusively on KB, whereas the bacteriophage lcI protein is an activator that normally exerts its effect exclusively on kf (4–6). CRP activates transcription from the lac promoter by binding to a recognition site centered 61.5 bp upstream from the start point of transcription and contacting the a subunit of RNAP (7). The a subunit consists of two independently folded domains, an N- terminal domain (NTD) and a C-terminal domain (CTD), sepa- rated by a flexible linker region (8–10). Whereas the aNTD mediates formation of the a dimer and serves as a scaffold for the assembly of RNAP, the aCTD is a DNA-binding domain that also serves as the target for many transcriptional activators (1, 3, 11). When bound at the lac promoter, CRP has been shown to stabilize the binding of the aCTD to the DNA in the region between the CRP recognition site and the promoter 235 element (7). Thus, CRP appears to work by a simple cooperative binding mechanism (1–3), stabilizing the closed complex at the lac promoter (4, 6). In contrast, lcI activates transcription from the l promoter PRM when bound to an operator site centered 42 bp upstream from the start point of transcription and is thought to contact the s subunit of RNAP (reviewed in ref. 12). lcI is a two-domain protein that binds as a dimer to its operator sites on the phage chromosome (13). Its NTD contains a helix-turn-helix DNA-binding motif, and its CTD mediates dimer formation as well as cooperative binding to pairs of operator sites (14). At the right operator region (OR), lcI dimers bind cooperatively to sites OR1 and OR2, and the dimer at OR2 activates transcription from promoter PRM (see Fig. 1A) (13). The isolation of lcI mutants specifically defective for activation (positive control mutants) led originally to the identification of a positive control surface located in the NTD of lcI (15–17). The suggestion that lcI uses this positive control surface to contact the s subunit of RNAP is based on the isolation and analysis of s mutants that affect lcI-stimulated transcription (18–20). The s subunit of RNAP is responsible for recognition of specific promoter sequences; alternative s factors combine with the enzy- matic core (a2bb9) to form alternative holoenzyme species (21). Promoter PRM is recognized by the s70 form of RNAP (Es70), which directs transcription of the majority of E. coli genes. The s70 subunit makes base-specific contacts with the promoter in both its 210 and 235 regions, using conserved regions 2 and 4, respectively, to do so (see ref. 21). lOR2 is centered just upstream of the 235 region of PRM (at position 242), and residues in region 4 of s70, which contains a putative helix-turn-helix DNA-binding motif, have been implicated in the interaction with lcI (18, 19). Nevertheless, there has been no direct demonstration of an interaction between lcI and region 4 of s70. Here, we design an in vivo assay that permits the detection of an energetically favorable interaction between lcI and a s fragment encompassing region 4. Specifically, we tether the relevant portion of s to the a subunit of RNAP and then show that lcI can activate transcription from a suitably designed test promoter by stabilizing the binding of the transplanted s moiety to an ectopic 235 element (Fig. 1B). We present a model that explains how the ability of lcI to stabilize the binding of region 4 of s to the DNA can account for its effect on the rate of isomerization at PRM. Materials and Methods Plasmids and Strains. Plasmid pAClcI harbors the wild-type cI gene under the control of the lacUV5 promoter (22). pAClcI(Sa109) is a derivative of pAClcI and encodes lcI(S35L, D38Y, K39N). pAClcI(Sa104) is a derivative of pAClcI and encodes lcI(S35L, D38Y, K39E). Plasmids pAClcI(Sa109) and pAClcI(Sa104) were constructed by cloning the appropriate NdeI–NsiI cut PCR prod- ucts [made using plasmids pFB109 and pFB104 (see ref. 17) as templates] into an NdeI–NsiI cut derivative of pAClcI that contains an NdeI site at the start of the cI gene. pAClcI(Sa109-Y38N) is a derivative of pAClcI(Sa109) in which the Y38N change was introduced by the PCR. pAClcI(D38N) is a derivative of pAClcI that was made by cloning an NdeI–NsiI cut PCR product [made using plasmid p16 (see ref. 16) as template] into the NdeI–NsiI cut This paper was submitted directly (Track II) to the PNAS office. Abbreviations: OR, operator region; RNAP, RNA polymerase; CRP, cAMP receptor protein; NTD, N-terminal domain; CTD, C-terminal domain. *To whom reprint requests should be addressed. E-mail: ahochschild@hms.harvard.edu. The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact. PNAS u November 21, 2000 u vol. 97 u no. 24 u 13215–13220 G EN ET IC S Mechanism for a transcriptional activator that works at the isomerization step Simon L. Dove, Franklin W. Huang, and Ann Hochschild* Department of Microbiology and Molecular Genetics, Harvard Medical School, Boston, MA 02115 Edited by Mark Ptashne, Memorial Sloan-Kettering Cancer Center, New York, NY, and approved September 26, 2000 (received for review June 30, 2000) Transcriptional activators in prokaryotes have been shown to stimulate different steps in the initiation process including the initial binding of RNA polymerase (RNAP) to the promoter and a postbinding step known as the isomerization step. Evidence sug- gests that activators that affect initial binding can work by a cooperative binding mechanism by making energetically favorable contacts with RNAP, but the mechanism by which activators affect the isomerization step is unclear. A well-studied example of an activator that normally exerts its effect exclusively on the isomer- ization step is the bacteriophage A cl protein (Acl), which has been shown genetically to interact with the C-terminal region of the o7° subunit of RNAP. We show here that the interaction between Acl and o can stimulate transcription even when the relevant portion of o is transplanted to another subunit of RNAP. This activation depends on the ability of Acl to stabilize the binding of the transplanted o moiety to an ectopic —35 element. Based on these and previous findings, we discuss a simple model that explains how an activator’s ability to stabilize the binding of an RNAP subdomain to the DNA can account for its effect on either the initial binding of RNAP to a promoter or the isomerization step. M any transcriptional activators in prokaryotes bind to spe- cific sequences associated with the promoters they regu- late and affect the initiation process through direct contacts with RNA polymerase (RNAP; subunit structure, a2BB'a) (1-3). The process of transcription initiation in Escherichia coli can be described by a simplified two-step model (4). First, RNAP binds to fully duplex promoter DNA to form what is called the closed complex. Formation of this complex is reversible and is described by an equilibrium binding constant Kg. For transcription to initiate, the closed complex must then isomerize to form the transcriptionally active open complex in which the DNA is locally melted to expose the transcription start site. This isomerization step is usually irreversible and is described by a forward rate constant ke. The cAMP receptor protein (CRP) is a well- characterized example of an activator that can exert its effect exclusively on Kp, whereas the bacteriophage AcI protein is an activator that normally exerts its effect exclusively on k¢ (4-6). CRP activates transcription from the Jac promoter by binding to a recognition site centered 61.5 bp upstream from the start point of transcription and contacting the a subunit of RNAP (7). The a subunit consists of two independently folded domains, an N- terminal domain (NTD) and a C-terminal domain (CTD), sepa- rated by a flexible linker region (8-10). Whereas the aNTD mediates formation of the a dimer and serves as a scaffold for the assembly of RNAP, the aCTD is a DNA-binding domain that also serves as the target for many transcriptional activators (1, 3, 11). When bound at the /Jac promoter, CRP has been shown to stabilize the binding of the aCTD to the DNA in the region between the CRP recognition site and the promoter —35 element (7). Thus, CRP appears to work by a simple cooperative binding mechanism (1-3), stabilizing the closed complex at the Jac promoter (4, 6). In contrast, AcI activates transcription from the A promoter Pam when bound to an operator site centered 42 bp upstream from the start point of transcription and is thought to contact the o subunit of RNAP (reviewed in ref. 12). AcI is a two-domain protein that binds as a dimer to its operator sites on the phage chromosome (13). Its NTD contains a helix-turn-helix DNA-binding motif, and its CTD mediates dimer formation as well as cooperative binding to pairs of operator sites (14). At the right operator region (Or), AcI dimers bind cooperatively to sites Op1 and Op2, and the dimer at Op? activates transcription from promoter Prm (see Fig. 1A) (13). The isolation of AcI mutants specifically defective for activation (positive control mutants) led originally to the identification of a positive control surface located in the NTD of AcI (15-17). The suggestion that AcI uses this positive control surface to contact the o subunit of RNAP is based on the isolation and analysis of a mutants that affect AcI-stimulated transcription (18-20). The o subunit of RNAP is responsible for recognition of specific promoter sequences; alternative o factors combine with the enzy- matic core (a2B8B’) to form alternative holoenzyme species (21). Promoter Pam is recognized by the o” form of RNAP (Eo”), which directs transcription of the majority of E. coli genes. The o”° subunit makes base-specific contacts with the promoter in both its —10 and —35 regions, using conserved regions 2 and 4, respectively, to do so (see ref. 21). AOg2 is centered just upstream of the —35 region of Pm (at position —42), and residues in region 4 of o”°, which contains a putative helix-turn-helix DNA-binding motif, have been implicated in the interaction with AcI (18, 19). Nevertheless, there has been no direct demonstration of an interaction between Acl and region 4 of 07”. Here, we design an in vivo assay that permits the detection of an energetically favorable interaction between AcI and a o fragment encompassing region 4. Specifically, we tether the relevant portion of o to the a subunit of RNAP and then show that AcI can activate transcription from a suitably designed test promoter by stabilizing the binding of the transplanted o moiety to an ectopic —35 element (Fig. 1B). We present a model that explains how the ability of AcI to stabilize the binding of region 4 of o to the DNA can account for its effect on the rate of isomerization at Pew. Materials and Methods Plasmids and Strains. Plasmid pACAcI harbors the wild-type c/ gene under the control of the JacUV5 promoter (22). pACAcI(Sa109) is a derivative of pACAcI and encodes AcI(S35L, D38Y, K39N). pACAcI(Sa104) is a derivative of pACAcI and encodes AcI(S35L, D38Y, K39E). Plasmids pACAcI(Sa109) and pACAcI(Sa104) were constructed by cloning the appropriate NdeI-NsiI cut PCR prod- ucts [made using plasmids pFB109 and pFB104 (see ref. 17) as templates] into an NdeI-NsiI cut derivative of pACAcI that contains an NdeI site at the start of the cJ gene. pACAcI(Sa109-Y38N) is a derivative of pACAcI(Sal09) in which the Y38N change was introduced by the PCR. pACAcI(D38N) is a derivative of pACAcI that was made by cloning an NdeI-Nsil cut PCR product [made using plasmid p16 (see ref. 16) as template] into the NdeI—NsiI cut This paper was submitted directly (Track II) to the PNAS office. Abbreviations: Or, operator region; RNAP, RNA polymerase; CRP, cAMP receptor protein; NTD, N-terminal domain; CTD, C-terminal domain. *To whom reprint requests should be addressed. E-mail: ahochschild@hms.harvard.edu. The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. § 1734 solely to indicate this fact. PNAS | November 21,2000 | vol.97 | no.24 | 13215-13220 GENETICS derivative of pAClcI. pACDcI and pBRa have been described previously (22). Plasmid pBRa-s70 encodes residues 1–248 of the a subunit of E. coli RNAP fused to residues 528–613 of the s70 subunit of E. coli RNAP under the control of tandem lpp and lacUV5 pro- moters. The hybrid a-s70 gene was created using overlap PCR and cloned into EcoRI–BamHI-digested pBRa to make pBRa- s70. Plasmid pBRa-s70(R596H) was made in a similar manner and is identical to pBRa-s70, except the s moiety of the encoded chimera contains the R596H substitution. pBRa-s70(R588H) is a derivative of pBRa-s70 in which the R588H change in the s moiety of the chimera was introduced by the PCR. Plasmid pBRa-s38 encodes residues 1–248 of the a subunit of E. coli RNAP fused to residues 243–330 of the s38 subunit of E. coli RNAP under the control of tandem lpp and lacUV5 promoters. pBRa-s38 was made essentially the same way as pBRa-s70. Plasmid pFW11-OR2–55yCons-35 was constructed by cloning an EcoRI–HindIII cut PCR product containing the lac promoter derivative plac OR2–55yCons-35 into pFW11 (23) cut with EcoRI– HindIII. pFW11-OR2–55yCons-35 was then transformed into strain CSH100, and the promoter-lacZ fusion was recombined onto an F9 episome and mated into strain FW102 to create reporter strain SF1 (see ref. 23). Plasmid pFW11-OR2–55yTTAACA was similarly constructed, and reporter strain SF2, which is identical to strain SF1 except for the sequence of the auxiliary 235 element of the test promoter (TTAACA instead of TTGACA), was made in the same manner as strain SF1. The PCR-amplified regions of all constructs were sequenced to confirm that no errors had been introduced as a result of the PCR process. Experimental Procedures For all experiments, cells were grown in LB supplemented with carbenicillin (50 mg ml21), chloramphenicol (25 mg ml21), and kanamycin (50 mg ml21) together with isopropyl-b-D-thiogalacto- side (IPTG) at the concentration indicated. SDS-CHCl3 perme- abilized cells were assayed for b-galactosidase activity essentially as described (24). Assays were done at least three times in duplicate on separate occasions, with similar results. Values are the averages from one experiment; duplicate measurements differed by ,10%. For all primer extension analyses, IPTG was added to the growth medium to a final concentration of 50 mM. RNA isolation, primer labeling, and primer extension assays were essentially as described previously (22). Results Design of the Experiment. We devised a strategy for assaying the ability of lcI to interact with region 4 of s in vivo. This strategy was based on our previous demonstration that transcription can be activated by any sufficiently strong contact between a DNA- bound protein and a protein domain fused to RNAP (refs. 22 and 24; see also refs. 2 and 25). In particular, we showed that protein domains fused to the a subunit of RNAP in place of the aCTD can mediate transcriptional activation by serving as artificial activation targets for DNA-bound proteins (22, 24). We also showed that transcription can be activated by a sufficiently strong protein–DNA interaction between a DNA-binding domain teth- ered to RNAP and a cognate recognition site positioned up- stream of a test promoter (24). Accordingly, we reasoned that region 4 of s70 might be able to activate transcription from a suitably designed test promoter (bearing an auxiliary 235 ele- ment) when tethered to the aNTD. We anticipated further that such a system should allow us to detect energetically favorable protein–protein interactions between the tethered s moiety and adjacently bound proteins because such interactions would sta- bilize the binding of the s moiety to the DNA and hence increase the magnitude of the activation. Following this strategy, we replaced the aCTD with a C- terminal fragment of s70 encompassing region 4 and constructed a test promoter bearing an auxiliary 235 element in the up- stream region together with a flanking l operator. This exper- imental setup enabled us to ask whether a DNA-bound lcI dimer would activate transcription from the test promoter by stabilizing the binding of the tethered s moiety to the auxiliary 235 element (see Fig. 1B). The hybrid a2s70 gene consisted of codons 1–248 of a fused to codons 528–613 of s70 (the final 86 codons). The test promoter plac OR2–55yCons-35 consisted of the lac core promoter, a second 235 hexamer centered at position 245.5, and the flanking l operator (OR2) centered at position 255. The position of OR2 relative to the auxiliary 235 element is the same as at the l PRM promoter. lcI Proteins Activate Transcription from the Test Promoter in the Presence of the a-s Chimera. The test promoter (plac OR2–55y Cons-35) was fused to the lacZ gene and introduced into E. coli strain FW102 (23) in single copy on an F9 episome to create reporter strain SF1. We assayed the abilities of wild-type lcI and two superactivating variants (17) to activate transcription from the test promoter in the presence or absence of the a-s70 chimera. Superactivators 104 and 109 activated transcription a maximum of '6-fold in the presence of the a-s70 chimera, and wild-type lcI activated transcription weakly (,2-fold) (Fig. 2A). This difference in the stimulatory activities of wild-type lcI and the superactivators mirrors that previously observed when the proteins were assayed for their abilities to activate transcription from l PRM (17). The lcI proteins only activated transcription from the test promoter in the presence of the a-s70 chimera; no stimulation was detected in the presence of excess wild-type a (Fig. 2 A) or excess wild-type s70 (data not shown). Furthermore, the a-s70 chimera did not mediate any stimulatory effect in the absence of lcI (data not shown). Primer extension analysis confirmed that the three lcI proteins stimulated the production of correctly initiated transcripts (Fig. 2B). Transcriptional Activation by lcI Proteins at the Test Promoter De- pends on the Activating Region of lcI and on the Ability of the Tethered s Moiety to Interact with the Auxiliary 235 Element. The results shown in Fig. 2 suggest that lcI and the two superacti- vators are interacting with the tethered s moiety and stabilizing its binding to the auxiliary 235 element. To test the hypothesis that the observed activation depends on the positive control surface of lcI, we introduced a positive control mutation into the Fig. 1. (A) lcI binds cooperatively to operator sites OR1 and OR2 to activate transcription from PRM. Activation is mediated by the lcI dimer bound at OR2, which likely contacts the s70 subunit of RNAP. (B) Genetic strategy for detecting the interaction between lcI and region 4 of s70. Replacement of the RNAP aCTD with region 4 of s70 permits interaction between the transplanted region of s70 and a lcI dimer bound adjacent to an auxiliary 235 element. The artificial promoter derivative plac OR2–55yCons-35 is shown; this bears the auxiliary 235 element (TTGACA) and the l operator OR2, centered 45.5 bp and 55 bp, respec- tively, upstream of the transcriptional start site of the lac promoter. 13216 u www.pnas.org Dove et al. 6 region 4 Fig. 1. (A) Acl binds cooperatively to operator sites Og1 and Op2 to activate transcription from Pam. Activation is mediated by the Acl dimer bound at Or2, which likely contacts the o7° subunit of RNAP. (B) Genetic strategy for detecting the interaction between Acl and region 4 of «7°. Replacement of the RNAP aCTD with region 4 of 7° permits interaction between the transplanted region of «7° and a Acl dimer bound adjacent to an auxiliary —35 element. The artificial promoter derivative plac Og2-55/Cons-35 is shown; this bears the auxiliary —35 element (TTGACA) and the A operator Op2, centered 45.5 bp and 55 bp, respec- tively, upstream of the transcriptional start site of the /ac promoter. derivative of pACAcI. pACAcI and pBRa have been described previously (22). Plasmid pBRa-o” encodes residues 1-248 of the a subunit of E. coli RNAP fused to residues 528-613 of the 0” subunit of E. coli RNAP under the control of tandem /pp and lacUVS5 pro- moters. The hybrid a-o” gene was created using overlap PCR and cloned into EcoRI-BamHlI-digested pBRa to make pBRa- a”, Plasmid pBRa-o7°(R596H) was made in a similar manner and is identical to pBRa-o”, except the o moiety of the encoded chimera contains the R596H substitution. pBRa-o7°(R588H) is a derivative of pBRa-o” in which the R588H change in the moiety of the chimera was introduced by the PCR. Plasmid pBRa-o*® encodes residues 1-248 of the a subunit of E. coli RNAP fused to residues 243-330 of the o°® subunit of E. coli RNAP under the control of tandem /pp and lacUV5 promoters. pBRa-o*® was made essentially the same way as pBRa-o”. Plasmid pFW11-Og2-55 /Cons-35 was constructed by cloning an EcoRI-Hindlll cut PCR product containing the Jac promoter derivative plac Og2-55/Cons-35 into pFW11 (23) cut with EcoRI- Hind. pFW11-Op2-55/Cons-35 was then transformed into strain CSH100, and the promoter-/acZ fusion was recombined onto an F’ episome and mated into strain FW102 to create reporter strain SF1 (see ref. 23). Plasmid pFW11-Og2-55/TTAACA was similarly constructed, and reporter strain SF2, which is identical to strain SF1 except for the sequence of the auxiliary —35 element of the test promoter (TTAACA instead of TTGACA), was made in the same manner as strain SF1. The PCR-amplified regions of all constructs were sequenced to confirm that no errors had been introduced as a result of the PCR process. Experimental Procedures For all experiments, cells were grown in LB supplemented with carbenicillin (50 wg ml~!), chloramphenicol (25 wg ml~!), and kanamycin (50 yg ml~') together with isopropyl-8-D-thiogalacto- side (IPTG) at the concentration indicated. SDS-CHCl; perme- abilized cells were assayed for B-galactosidase activity essentially as described (24). Assays were done at least three times in duplicate on separate occasions, with similar results. Values are the averages from one experiment; duplicate measurements differed by <10%. For all primer extension analyses, IPTG was added to the growth medium to a final concentration of 50 wM. RNA isolation, primer labeling, and primer extension assays were essentially as described previously (22). 13216 | www.pnas.org Results Design of the Experiment. We devised a strategy for assaying the ability of AcI to interact with region 4 of a in vivo. This strategy was based on our previous demonstration that transcription can be activated by any sufficiently strong contact between a DNA- bound protein and a protein domain fused to RNAP (refs. 22 and 24; see also refs. 2 and 25). In particular, we showed that protein domains fused to the a subunit of RNAP in place of the aCTD can mediate transcriptional activation by serving as artificial activation targets for DNA-bound proteins (22, 24). We also showed that transcription can be activated by a sufficiently strong protein-DNA interaction between a DNA-binding domain teth- ered to RNAP and a cognate recognition site positioned up- stream of a test promoter (24). Accordingly, we reasoned that region 4 of o”° might be able to activate transcription from a suitably designed test promoter (bearing an auxiliary —35 ele- ment) when tethered to the aNTD. We anticipated further that such a system should allow us to detect energetically favorable protein-protein interactions between the tethered o moiety and adjacently bound proteins because such interactions would sta- bilize the binding of the a moiety to the DNA and hence increase the magnitude of the activation. Following this strategy, we replaced the aCTD with a C- terminal fragment of o”° encompassing region 4 and constructed a test promoter bearing an auxiliary —35 element in the up- stream region together with a flanking A operator. This exper- imental setup enabled us to ask whether a DNA-bound AcI dimer would activate transcription from the test promoter by stabilizing the binding of the tethered o moiety to the auxiliary —35 element (see Fig. 1B). The hybrid a—o” gene consisted of codons 1-248 of a fused to codons 528-613 of o7° (the final 86 codons). The test promoter plac Op2-55/Cons-35 consisted of the lac core promoter, a second —35 hexamer centered at position —45.5, and the flanking A operator (Op2) centered at position —55. The position of Og? relative to the auxiliary —35 element is the same as at the A Pym promoter. Acl Proteins Activate Transcription from the Test Promoter in the Presence of the a-o Chimera. The test promoter (plac Og2-55/ Cons-35) was fused to the JacZ gene and introduced into E. coli strain FW102 (23) in single copy on an F’ episome to create reporter strain SF1. We assayed the abilities of wild-type AcI and two superactivating variants (17) to activate transcription from the test promoter in the presence or absence of the a-c” chimera. Superactivators 104 and 109 activated transcription a maximum of ~6-fold in the presence of the a-o7° chimera, and wild-type AclI activated transcription weakly (<2-fold) (Fig. 2A). This difference in the stimulatory activities of wild-type AcI and the superactivators mirrors that previously observed when the proteins were assayed for their abilities to activate transcription from A Pro (17). The AcI proteins only activated transcription from the test promoter in the presence of the a-o” chimera; no stimulation was detected in the presence of excess wild-type a (Fig. 2A) or excess wild-type o”° (data not shown). Furthermore, the a-o” chimera did not mediate any stimulatory effect in the absence of AcI (data not shown). Primer extension analysis confirmed that the three AcI proteins stimulated the production of correctly initiated transcripts (Fig. 2B). Transcriptional Activation by Acl Proteins at the Test Promoter De- pends on the Activating Region of AcI and on the Ability of the Tethered o Moiety to Interact with the Auxiliary —35 Element. The results shown in Fig. 2 suggest that AcI and the two superacti- vators are interacting with the tethered o moiety and stabilizing its binding to the auxiliary —35 element. To test the hypothesis that the observed activation depends on the positive control surface of AcI, we introduced a positive control mutation into the Dove et al. genes encoding lcISa109 and lcISa104. The resulting lcI vari- ants (bearing amino acid substitution Y38N) manifested a substantially decreased ability to activate transcription from the test promoter (Fig. 3B and data not shown). We confirmed that these activation defects were not attributable to DNA-binding defects of the altered superactivators by performing an in vivo repression assay using a test promoter bearing a single l operator between its 210 and 235 regions (data not shown). We then tested the hypothesis that the observed activation depends on the ability of the tethered s moiety to interact with the auxiliary 235 element positioned adjacent to the l operator. To do this, we introduced mutations predicted to disrupt this interaction into either the 235 element (see Fig. 3A) or the gene encoding the a-s70 chimera. When we weakened the auxiliary 235 element present on the reporter template by introducing a G to A substi- tution at the third position, lcISa109 failed to activate transcription from the test promoter (Fig. 3B). Similarly, lcISa109 failed to activate transcription in the presence of a mutant a-s70 chimera bearing an amino acid substitution in the s moiety (R588H) predicted to disrupt DNA binding (26) (Fig. 3B). Activation with a Mutant–Suppressor Pair. Our results demonstrate that wild-type lcI and two superactivating variants can stabilize the sequence-specific binding of a C-terminal fragment of s70 to DNA. For wild-type lcI, this effect is close to the threshold of detection in our in vivo assay, and for the superactivating variants the effects are larger, as predicted based on their activities at PRM (17). Because these superactivating variants of lcI have not been sub- jected to kinetic analysis at PRM, we examined the effect of another lcI mutant that has been analyzed kinetically and was found to stimulate the rate of isomerization more efficiently than wild-type lcI. Mutant lcI-D38N is a positive control mutant (16), the activation defect of which can be suppressed by a mutant form of s bearing the substitution R596H (s-R596H) (18). In vitro exper- iments done with reconstituted mutant RNAP (Es-R596H) re- vealed that lcI-D38N exerts its effect exclusively on the isomer- ization step, producing a 27-fold increase in the isomerization rate constant (kf) as compared with a 17-fold increase produced by wild-type lcI working on wild-type RNAP (20). To test whether our artificial system could detect an interac- tion between lcI-D38N and s-R596H, we compared the abilities of lcI-D38N to activate transcription from the artificial pro- moter in the presence of the a-s70 chimera with or without the R596H substitution in the s moiety. lcI-D38N failed to activate transcription in the presence of the unmodified form of the a-s70 chimera but activated transcription '4-fold when the s moiety of the chimera bore the R596H substitution (Fig. 4). Primer extension analysis confirmed that this activation reflected an increase in correctly initiated transcripts (data not shown). Like the unmodified form of the a-s70 chimera, the a-s70 (R596H) variant did not activate transcription from the test promoter in the absence of lcI-D38N (data not shown). Superactivating Variants of lcI Activate Transcription from the Test Promoter in the Presence of an a-s38 Chimera. The stationary phase s factor, s38, is very similar to s70 in the DNA-binding regions, particularly throughout region 4.2, and recognizes the same 235 Fig. 2. Effects of wild-type lcI and lcI superactivators on transcription in the presence of the a-s70 chimera. (A) SF1 cells harboring the indicated plasmids were assayed for b-galactosidase activity. pACYC-derived plasmids encoded lcI (pAClcI), lcISa109 [pAClcI(Sa109)], or lcISa104 [pAClcI(Sa104)]; pBR322- derived plasmids encoded either the a-s70 chimera (pBRa-s70) or wild-type a (pBRa). (B) Primer extension analysis of transcripts produced from plac OR2– 55yCons-35 with wild-type lcI or lcI superactivators in the presence of the a-s70 chimera. Total RNA was isolated from SF1 cells harboring plasmids encoding the indicated proteins, and the primer extension analysis was done by using a primer complementary to the lacZ transcript produced by the plac OR2–55yCons-35 promoter. Primer extension products produced by correctly initiated transcripts are indicated by 11. Excess unincorporated primer is shown in the lower panel. Fig. 3. Effects of mutations in the a-s70 chimera or the additional 235 element on activation by lcI superactivator. (A) Schematic of test promoters used to determine the effect of mutating the additional 235 element on transcriptional activation by Sa109 in the presence of the a-s70 chimera. The sequences of the additional 235 elements from the two test promoters are indicated (consensus 5 TTGACA). (B) SF1 and SF2 cells harboring the indicated plasmids were assayed for b-galactosidase activity. In each panel, the sequence of the additional 235 element from the relevant test promoter is given. pACYC-derived plasmids en- coded either lcISa109 [pAClcI(Sa109)] or lcISa109-Y38N [pAClcI(Sa109-Y38N)]; pBR322-derived plasmids encoded the a-s70 chimera (pBRa-s70), the a-s70 (R588H) chimera [pBRa-s70 (R588H)], or wild-type a (pBRa). Dove et al. PNAS u November 21, 2000 u vol. 97 u no. 24 u 13217 G EN ET IC S > 500 pAC AcI(Sa109) + pBRa-o”? 400 PAC AcI(Sa104) + pBRa-o” 300 200 100 PAC AcI + pBRa-o7 T T J 10 20 30 40 50 B-Galactosidase activity (Miller units) o 200 100 __ PAC Acl(Sa104) + pBRa 7 v Ene elt Sa + pBRa oT T T T T 0 10 20 30 40 50 IPTG (uM) 6 es, es rm: PRP SF HS GO" shehiothoe oe Sea oF +1 we ot Fig.2. Effects of wild-type Acl and Acl superactivators on transcription in the presence of the a-a”° chimera. (A) SF1 cells harboring the indicated plasmids were assayed for B-galactosidase activity. pACYC-derived plasmids encoded Acl (pACAcl), AcISa109 [pACAcI(Sa109)], or AcISa104 [pACAcI(Sa104)]; pBR322- derived plasmids encoded either the a-a”° chimera (pBRa-o”°) or wild-type a (pBRa). (B) Primer extension analysis of transcripts produced from plac Og2- 55/Cons-35 with wild-type Acl or Acl superactivators in the presence of the a-o7° chimera. Total RNA was isolated from SF1 cells harboring plasmids encoding the indicated proteins, and the primer extension analysis was done by using a primer complementary to the /acZ transcript produced by the plac Or2-55/Cons-35 promoter. Primer extension products produced by correctly initiated transcripts are indicated by +1. Excess unincorporated primer is shown in the lower panel. genes encoding AcISa109 and AcISa104. The resulting AcI vari- ants (bearing amino acid substitution Y38N) manifested a substantially decreased ability to activate transcription from the test promoter (Fig. 3B and data not shown). We confirmed that these activation defects were not attributable to DNA-binding defects of the altered superactivators by performing an in vivo repression assay using a test promoter bearing a single A operator between its —10 and —35 regions (data not shown). We then tested the hypothesis that the observed activation depends on the ability of the tethered o moiety to interact with the auxiliary —35 element positioned adjacent to the A operator. To do this, we introduced mutations predicted to disrupt this interaction into either the —35 element (see Fig. 34) or the gene encoding the -o”° chimera. When we weakened the auxiliary —35 element present on the reporter template by introducing a G to A substi- tution at the third position, AcISa109 failed to activate transcription from the test promoter (Fig. 3B). Similarly, AcISa109 failed to activate transcription in the presence of a mutant a-o”° chimera bearing an amino acid substitution in the o moiety (R588H) predicted to disrupt DNA binding (26) (Fig. 3B). Activation with a Mutant-Suppressor Pair. Our results demonstrate that wild-type AcI and two superactivating variants can stabilize the Dove et al. A AOR2 = 738 ier} “19 Place On2-55 TTGACA TTAACA B 5009 TTGACA pAC Acl(Sa109) + pBRa-o”? 400 2 200 F 4 = 200 2 3 100 pAC AcI(Sa109-Y38N) + pBRa-o' g pAC Aci(Sa109) + pBRa- PuNSe8H) 3 O-t T T T T 1 ES © 110 2 30 40 50 @ 2004 TTAACA a 100 + __. pAC Acl(Sa109) + pBRa-o”? — PAC AcI(Sa109) + pBRa i?) T T T T 1 0 10 20 30 40 50 IPTG (uM) Fig.3. Effects of mutations in the a-o”° chimera or the additional —35 element on activation by Acl superactivator. (A) Schematic of test promoters used to determine the effect of mutating the additional —35 element on transcriptional activation by Sa109 in the presence of the a-a”° chimera. The sequences of the additional —35 elements from the two test promoters are indicated (consensus = TTGACA). (B) SF1 and SF2 cells harboring the indicated plasmids were assayed for B-galactosidase activity. In each panel, the sequence of the additional —35 element from the relevant test promoter is given. pACYC-derived plasmids en- coded either AcISa109 [pACAcI(Sa109)] or AcISa109-Y38N [pACAcl(Sa109-Y38N)]; pBR322-derived plasmids encoded the a-o”° chimera (pBRa-c”°), the a-c7® (R588H) chimera [pBRa-o7° (R588H)], or wild-type a (pBRa). sequence-specific binding of a C-terminal fragment of 0” to DNA. For wild-type AcI, this effect is close to the threshold of detection in our in vivo assay, and for the superactivating variants the effects are larger, as predicted based on their activities at Pam (17). Because these superactivating variants of AcI have not been sub- jected to kinetic analysis at Ppm, we examined the effect of another AcI mutant that has been analyzed kinetically and was found to stimulate the rate of isomerization more efficiently than wild-type AcI. Mutant AcI-D38N is a positive control mutant (16), the activation defect of which can be suppressed by a mutant form of o bearing the substitution R596H (o-R596H) (18). In vitro exper- iments done with reconstituted mutant RNAP (Eo-R596H) re- vealed that AcI-D38N exerts its effect exclusively on the isomer- ization step, producing a 27-fold increase in the isomerization rate constant (kp) as compared with a 17-fold increase produced by wild-type AcI working on wild-type RNAP (20). To test whether our artificial system could detect an interac- tion between AcI-D38N and o-R596H, we compared the abilities of AcI-D38N to activate transcription from the artificial pro- moter in the presence of the a-o” chimera with or without the R596H substitution in the o moiety. AcI-D38N failed to activate transcription in the presence of the unmodified form of the a-o” chimera but activated transcription ~4-fold when the o moiety of the chimera bore the R596H substitution (Fig. 4). Primer extension analysis confirmed that this activation reflected an increase in correctly initiated transcripts (data not shown). Like the unmodified form of the a-o” chimera, the a-o”? (R596H) variant did not activate transcription from the test promoter in the absence of AcI-D38N (data not shown). Superactivating Variants of Acl Activate Transcription from the Test Promoter in the Presence of an a-o?* Chimera. The stationary phase o factor, 0°, is very similar to o”° in the DNA-binding regions, particularly throughout region 4.2, and recognizes the same —35 PNAS | November 21,2000 | vol.97 | no.24 | 13217 GENETICS consensus sequence as does s70 (refs. 27 and 28; T. Gaal & R. L. Gourse, personal communication). We replaced the s70 moiety of the a-s70 chimera with the corresponding region of s38 (residues 243 to 330) and tested whether the resulting a-s38 chimera could mediate transcriptional activation from our arti- ficial test promoter. The experiment of Fig. 5A shows that both lcISa109 and lcISa104 stimulated transcription efficiently in the presence of the a-s38 chimera. Wild-type lcI appeared to exert a slight stimulatory effect on transcription in the presence of the a-s38 chimera (Fig. 5A); however, the interpretation of this effect is complicated by the results obtained in the absence of lcI (see Discussion). Fig. 5B shows that in the absence of any form of lcI, the a-s38 chimera activated transcription from the test promoter bearing the consen- sus (TTGACA) ectopic 235 element '6-fold. This activation evidently depends on the ability of the tethered s38 moiety to bind to the ectopic 235 element since replacement of the consensus element with a mutated element (TTAACA) significantly reduced the magnitude of the activation (Fig. 5B). Primer extension analysis confirmed that the activation mediated by the a-s38 chimera either in the absence or the presence of a lcI variant reflected an increase in correctly initiated transcripts (data not shown). Discussion Genetic Evidence That the Interaction Between lcI and the Tethered s Moiety Is the Same Interaction That Activates Transcription at PRM. Our demonstration that lcI can stabilize the binding of region 4 of s70 to a 235 element provides strong support for the idea that there is an energetically favorable interaction between the activating region of lcI and a complementary surface of s. Importantly, we observed a close correlation between the effects of amino acid substitutions in both lcI and region 4 of s70 on transcriptional activation at PRM (16–18) and at our artificial promoter. First, we showed that two superactivating variants of lcI (Sa104 and Sa109), so designated based on their behavior at PRM (17), also activated transcription more strongly than wild-type lcI at the artificial promoter. Second, we showed that activation by wild-type lcI and the superactivating variants at the artificial promoter was depen- dent on their having functional activating regions as defined by their abilities to activate transcription from PRM. Finally, we tested the effect of introducing into the s moiety of the a-s chimera an amino acid substitution (R596H) that suppresses the activation defect of a lcI positive control mutant (lcI-D38N) at PRM (18); in the context of the a-s chimera, this amino acid substitution specifically en- hanced the ability of lcI-D38N to activate transcription from the artificial promoter. Mechanism by Which lcI Influences the Isomerization Step at PRM. We have shown that an activator (either wild-type lcI or lcI-D38N), which is known to function by accelerating the rate of isomer- ization at PRM, can activate transcription from our artificial promoter by stabilizing the binding of a tethered s moiety encompassing region 4 to an ectopic 235 element. The simplest interpretation of these findings is that both wild-type lcI and lcI-D38N similarly stabilize the binding of intact s70 to the 235 element when they activate transcription from PRM. How can this proposal be reconciled with the observed kinetics of the activa- tion process? Any detailed mechanistic model for the action of lcI at PRM must account both for its stimulatory effect on kf and for its lack of an effect on the initial binding step (described by an equilibrium constant KB for the formation of the closed complex). We suggest that when Es70 (or Es70-R596H) forms a closed complex at PRM, the activating region of lcI (or lcI- D38N) and its target surface on s are improperly aligned so that no energetically significant interaction can occur (Fig. 6A). We propose further that during the transition from the closed to the transcriptionally active open (melted) complex, the activating region of lcI and its target on s come into alignment, thus permitting an energetically significant interaction to occur (Fig. 6A). To explain the stimulatory effect of lcI on the rate of isomerization, we postulate that lcI stabilizes an intermediate along the pathway from the closed to the open complex, the formation of which limits the rate of initiation. More particu- larly, we suggest that during the isomerization process, there may be a tendency for region 4 of s to disengage from the 235 element, which limits open complex formation (Fig. 6B). DNA- bound lcI would function to counteract this tendency, thus Fig. 4. Effects of wild-type and mutant lcI on transcription in the presence of a-s70 chimeras. SF1 cells harboring the indicated plasmids were assayed for b-galactosidase activity. pACYC-derived plasmids encoded either lcI (pAClcI) or lcI(D38N) [pAClcI(D38N)]; pBR322-derived plasmids encoded either the a-s70 chimera (pBRa-s70) or the a-s70 (R596H) chimera [pBRa-s70 (R596H)]. Fig. 5. (A) Effects of wild-type lcI and lcI superactivators on transcription in the presence of the a-s38 chimera. SF1 cells harboring the indicated plasmids were assayed for b-galactosidase activity. pACYC-derived plasmids encoded lcI (pAClcI), lcISa109 [pAClcI(Sa109)], or lcISa104 [pAClcI(Sa104)]; the pBR322-derived plasmid encoded the a-s38 chimera (pBRa-s38). (B) Interaction between s38 region 4 and the DNA mediates transcriptional activation. SF1 and SF2 cells harboring the indicated plasmids were assayed for b-galactosi- dase activity. In each panel, the sequence of the additional 235 element from the relevant test promoter is given. The pACYC-derived plasmid encoded no lcI (pACDcI); pBR322-derived plasmids encoded either the a-s38 chimera (pBRa-s38) or wild-type a (pBRa). 13218 u www.pnas.org Dove et al. _ 07 pAC AcI (D38N) + pBRa-o7? (R596H) cA) § 300-4 s = 250 4 P= S 20074 8 @ 150-4 8 pAC Acl + pBRa-o7° @ 100 ———o pAC AcI + pBRa-o7° (R596H) g . Hv ———¥ pAC Acl (D38N) + pBRa-o7 8 50 5 a 0 T +—_T—T~T—¥_T-T_T—1 0 10 #2 430 40 50 IPTG (uM) Fig. 4. Effects of wild-type and mutant Acl on transcription in the presence of a-o7° chimeras. SF1 cells harboring the indicated plasmids were assayed for B-galactosidase activity. pACYC-derived plasmids encoded either Acl (pACAcl) or Acl(D38N) [pACAcI(D38N)]; pBR322-derived plasmids encoded either the a-o7° chimera (pBRa-o”°) or the a-a7° (R596H) chimera [pBRa-o”° (R596H)]. consensus sequence as does o” (refs. 27 and 28; T. Gaal & R. L. Gourse, personal communication). We replaced the a” moiety of the a-c”? chimera with the corresponding region of o°® (residues 243 to 330) and tested whether the resulting a-o*® chimera could mediate transcriptional activation from our arti- ficial test promoter. The experiment of Fig. 54 shows that both AcISa109 and AcISa104 stimulated transcription efficiently in the presence of the a-o** chimera. Wild-type AcI appeared to exert a slight stimulatory effect on 7 AC AcI (Sa109) + pBRa-o38 1400 p. ( ) +p = Dy S 6 L 1000 + pAC Aci (Sa104) + pBRo-o38 800 5 600 5 400 5 4 pAC AcI + pBRa-o38 200 7 B-Galactosidase activity (Miller units) 0 rs T ie) 10 20 30 40 50 IPTG (uM) is) 4007 TTGACA PAC AcI + pBRa-o3® 300 4 200 + 100 2 pAC AcI + pBRa ot T T T T 1 0 10 20 30 40 50 200 7 TTAACA pAC AcI + pBRa-o38 100 4 ee pA AcI + pBRa 0 T T T T 1 0 10 20 30 40 ~~ 50 IPTG (uM) B-Gatactosidase activity (Miller units) Fig.5. (A) Effects of wild-type Acl and Acl superactivators on transcription in the presence of the a-o?8 chimera. SF1 cells harboring the indicated plasmids were assayed for B-galactosidase activity. pACYC-derived plasmids encoded Acl (pACAcl), AcISa109 [pACAcI($a109)], or AcISa104 [pACAcl(Sa104)]; the pBR322-derived plasmid encoded the a-o38 chimera (pBRa-o*°). (B) Interaction between o°8 region 4 and the DNA mediates transcriptional activation. SF1 and SF2 cells harboring the indicated plasmids were assayed for B-galactosi- dase activity. In each panel, the sequence of the additional —35 element from the relevant test promoter is given. The pACYC-derived plasmid encoded no Acl (pACAcl); pBR322-derived plasmids encoded either the a-o2® chimera (pBRa-o?8) or wild-type a (pBRa). 13218 | www.pnas.org transcription in the presence of the a-o** chimera (Fig. 5A); however, the interpretation of this effect is complicated by the results obtained in the absence of AcI (see Discussion). Fig. 5B shows that in the absence of any form of AcI, the a-o** chimera activated transcription from the test promoter bearing the consen- sus (TTGACA) ectopic —35 element ~6-fold. This activation evidently depends on the ability of the tethered o** moiety to bind to the ectopic —35 element since replacement of the consensus element with a mutated element (TTAACA) significantly reduced the magnitude of the activation (Fig. 5B). Primer extension analysis confirmed that the activation mediated by the a-o** chimera either in the absence or the presence of a AcI variant reflected an increase in correctly initiated transcripts (data not shown). Discussion Genetic Evidence That the Interaction Between Aci and the Tethered o Moiety Is the Same Interaction That Activates Transcription at Pru. Our demonstration that AcI can stabilize the binding of region 4 of a’ to a —35 element provides strong support for the idea that there is an energetically favorable interaction between the activating region of AcI and a complementary surface of o. Importantly, we observed a close correlation between the effects of amino acid substitutions in both AcI and region 4 of o” on transcriptional activation at Pr (16-18) and at our artificial promoter. First, we showed that two superactivating variants of AcI (Sa104 and Sa109), so designated based on their behavior at Prm (17), also activated transcription more strongly than wild-type AcI at the artificial promoter. Second, we showed that activation by wild-type AcI and the superactivating variants at the artificial promoter was depen- dent on their having functional activating regions as defined by their abilities to activate transcription from Pru. Finally, we tested the effect of introducing into the o moiety of the a-o chimera an amino acid substitution (R596H) that suppresses the activation defect of a AclI positive control mutant (AcI-D38N) at Pam (18); in the context of the a-o chimera, this amino acid substitution specifically en- hanced the ability of AcI-D38N to activate transcription from the artificial promoter. Mechanism by Which Acl Influences the Isomerization Step at Pam. We have shown that an activator (either wild-type AcI or AcI-D38N), which is known to function by accelerating the rate of isomer- ization at Pym, can activate transcription from our artificial promoter by stabilizing the binding of a tethered o moiety encompassing region 4 to an ectopic —35 element. The simplest interpretation of these findings is that both wild-type AcI and AcI-D38N similarly stabilize the binding of intact 0” to the —35 element when they activate transcription from Pam. How can this proposal be reconciled with the observed kinetics of the activa- tion process? Any detailed mechanistic model for the action of AcI at Pam must account both for its stimulatory effect on k, and for its lack of an effect on the initial binding step (described by an equilibrium constant Kg for the formation of the closed complex). We suggest that when Eo” (or Eo”°-R596H) forms a closed complex at Prm, the activating region of AcI (or Acl- D38N) and its target surface on o are improperly aligned so that no energetically significant interaction can occur (Fig. 64). We propose further that during the transition from the closed to the transcriptionally active open (melted) complex, the activating region of AcI and its target on o come into alignment, thus permitting an energetically significant interaction to occur (Fig. 6A). To explain the stimulatory effect of AcI on the rate of isomerization, we postulate that AcI stabilizes an intermediate along the pathway from the closed to the open complex, the formation of which limits the rate of initiation. More particu- larly, we suggest that during the isomerization process, there may be a tendency for region 4 of o to disengage from the —35 element, which limits open complex formation (Fig. 6B). DNA- bound AcI would function to counteract this tendency, thus Dove et al. stabilizing a productive intermediate along the pathway to open complex formation (Fig. 6A). The ability of lcI to stabilize the binding of the tethered s moiety to the DNA at our artificial promoter but not when RNAP forms a closed complex at PRM suggests that region 4 of s may be more constrained in its natural context in the holoenzyme than when it is tethered to the aNTD by a flexible linker region. We think it likely that the interaction between lcI and the s moiety of the a-s chimera functions to stabilize the closed complex (affects KB) at our artificial test promoter. Using the same core promoter, we have previously shown that transcription can be activated by any sufficiently strong contact between a DNA-bound protein and a protein domain tethered to a subunit of RNAP or between a DNA-binding domain tethered to RNAP and a cognate recognition site positioned upstream of the core promoter (22, 24). Furthermore, our results established a cor- relation between the strength of the protein–protein (or the protein–DNA) interaction and the magnitude of the activation (ref. 22; S.L.D & A.H., unpublished results). The simplest interpretation of our findings is that these arbitrarily selected protein–protein or protein–DNA interactions stabilize the bind- ing of RNAP to the promoter. In the experimental setup used here, transcriptional activation results from the combined effects of a relatively weak protein–protein interaction (between lcI and the tethered s moiety) and a relatively weak protein–DNA interaction (between the tethered s moiety and the auxiliary 235 element). We note that the role of lcI at this artificial promoter is formally analogous to the role of CRP at the natural lac promoter; that is, CRP interacts with the aCTD and stabilizes its association with the DNA in the region between the CRP recognition site and the promoter 235 element (1, 7). In this case, CRP has been shown to exert its effect exclusively on KB (6). Regardless of the kinetic effect of lcI on transcription at our artificial test promoter, two principal findings, namely that lcI can (i) interact productively with its target when that target is transplanted from the s subunit to the a subunit and (ii) stabilize the binding of the transplanted s fragment to an ectopic 235 element, suggest that the functional significance of the lcI-s interaction is simply that contacts between s region 4 and the 235 element of PRM are stabilized. A Common Mechanism for the Effects of Activators That Work at Different Steps in the Initiation Process. An implication of our results is that there need not be any fundamental difference between an activator that affects KB and an activator that affects kf, both merely requiring a surface that can interact with an accessible complementary surface on RNAP (2, 3, 29). The kinetic effect of a particular activator working at a particular promoter may instead depend on when during the initiation process the appropriate surfaces can interact (29). If the inter- action can take place while RNAP is in the closed complex, an effect on KB would be expected, whereas if the interaction can take place only after the closed to open transition has begun, then an effect on kf would be expected. Thus, the same protein– protein interaction between an activator and RNAP might, in principle, produce an effect on KB, kf, or both, depending on spatial constraints imposed by the promoter itself and on the arrangement of the activator-binding site(s). This picture of the activation process provides an explanation for an unexpected effect uncovered by the kinetic analysis of the lcI-D38NyEs70-R596H mutantysuppressor pair, namely that wild-type lcI at PRM stimulates closed complex formation (i.e., exerts its effect predominantly on KB) when assayed with Es70- R596H (20). We suggest that this apparent change in activation mechanism may simply reflect a subtle change in the geometry of the interaction so that lcI can interact productively with region 4 of s70 when Es70-R596H forms a closed complex (29). It is possible that other s70-dependent activators that exert their effects on the rate of isomerization may in some cases do so by stabilizing the interaction of region 4 of s with the 235 element. In the case we have described, this stabilization appar- ently occurs by a simple cooperative binding mechanism: the activator contacts region 4 of s directly. In principle, however, direct contact with this DNA-binding domain of s would not necessarily be required. Instead, contact with another region or subunit of RNAP might function indirectly to stabilize the interaction between s region 4 and the 235 element. Several s70-dependent activators have been shown to affect the rate of isomerization (5, 20, 30, 31). A particularly well-characterized example is provided by CRP, which possesses at least two distinct activating regions (AR1 and AR2). When bound at a so-called class II promoter, which bears a CRP recognition site that overlaps the promoter 235 region, CRP uses AR1 to contact the aCTD and AR2 to contact the aNTD, the former contact mediating an effect on KB and the latter an effect on kf (31). Moreover, it has been shown that AR2 of CRP participates in an energetically favorable interaction with RNAP (31). Activators of another class have been shown to exert their effects on the isomerization step; these activators work on promoters recognized by the alternative s factor, s54, which appears to be unrelated to the s70 class of s factors (27). Unlike Fig. 6. Model for mechanism of action of lcI at PRM. (A) Activating region and its target (red patches) are misaligned in the closed complex but come into alignment subsequently during the process of open complex formation. Depicted in brackets is a hypothetical productive intermediate that is stabilized by lcI. (B) In the absence of lcI, formation of an unproductive intermediate limits open complex formation at PRM. Dove et al. PNAS u November 21, 2000 u vol. 97 u no. 24 u 13219 G EN ET IC S Closed Complex ——=> Closed Complex —=> Fig. 6. Productive Intermediate Stabilized by AcI Unproductive Intermediate ==> Open Complex Model for mechanism of action of Acl at Pram. (A) Activating region and its target (red patches) are misaligned in the closed complex but come into alignment subsequently during the process of open complex formation. Depicted in brackets is a hypothetical productive intermediate that is stabilized by Acl. (B) In the absence of Acl, formation of an unproductive intermediate limits open complex formation at Pam. stabilizing a productive intermediate along the pathway to open complex formation (Fig. 6A). The ability of AcI to stabilize the binding of the tethered o moiety to the DNA at our artificial promoter but not when RNAP forms a closed complex at Pram suggests that region 4 of o may be more constrained in its natural context in the holoenzyme than when it is tethered to the aNTD by a flexible linker region. We think it likely that the interaction between AcI and the o moiety of the a-o chimera functions to stabilize the closed complex (affects Kg) at our artificial test promoter. Using the same core promoter, we have previously shown that transcription can be activated by any sufficiently strong contact between a DNA-bound protein and a protein domain tethered to a subunit of RNAP or between a DNA-binding domain tethered to RNAP and a cognate recognition site positioned upstream of the core promoter (22, 24). Furthermore, our results established a cor- relation between the strength of the protein-protein (or the protein-DNA) interaction and the magnitude of the activation (ref. 22; S.L.D & A.H., unpublished results). The simplest interpretation of our findings is that these arbitrarily selected protein-protein or protein-DNA interactions stabilize the bind- ing of RNAP to the promoter. In the experimental setup used here, transcriptional activation results from the combined effects of a relatively weak protein-protein interaction (between AcI and the tethered o moiety) and a relatively weak protein-DNA interaction (between the tethered o moiety and the auxiliary —35 element). We note that the role of AcI at this artificial promoter is formally analogous to the role of CRP at the natural lac promoter; that is, CRP interacts with the aCTD and stabilizes its association with the DNA in the region between the CRP recognition site and the promoter —35 element (1, 7). In this case, CRP has been shown to exert its effect exclusively on Kg (6). Regardless of the kinetic effect of AcI on transcription at our artificial test promoter, two principal findings, namely that AcI can (i) interact productively with its target when that target is transplanted from the o subunit to the a subunit and (i) stabilize the binding of the transplanted o fragment to an ectopic —35 element, suggest that the functional significance of the AcI-o interaction is simply that contacts between o region 4 and the —35 element of Pry are stabilized. A Common Mechanism for the Effects of Activators That Work at Different Steps in the Initiation Process. An implication of our results is that there need not be any fundamental difference between an activator that affects Kg and an activator that affects Dove et al. ks, both merely requiring a surface that can interact with an accessible complementary surface on RNAP (2, 3, 29). The kinetic effect of a particular activator working at a particular promoter may instead depend on when during the initiation process the appropriate surfaces can interact (29). If the inter- action can take place while RNAP is in the closed complex, an effect on Kg would be expected, whereas if the interaction can take place only after the closed to open transition has begun, then an effect on k¢ would be expected. Thus, the same protein-— protein interaction between an activator and RNAP might, in principle, produce an effect on Kp, k,, or both, depending on spatial constraints imposed by the promoter itself and on the arrangement of the activator-binding site(s). This picture of the activation process provides an explanation for an unexpected effect uncovered by the kinetic analysis of the AcI-D38N/Eo7°-R596H mutant/suppressor pair, namely that wild-type AcI at Pam stimulates closed complex formation (i.e., exerts its effect predominantly on Ky) when assayed with Eo”?- R596H (20). We suggest that this apparent change in activation mechanism may simply reflect a subtle change in the geometry of the interaction so that AcI can interact productively with region 4 of o”° when Eo”°-R596H forms a closed complex (29). It is possible that other o7°-dependent activators that exert their effects on the rate of isomerization may in some cases do so by stabilizing the interaction of region 4 of o with the —35 element. In the case we have described, this stabilization appar- ently occurs by a simple cooperative binding mechanism: the activator contacts region 4 of o directly. In principle, however, direct contact with this DNA-binding domain of o would not necessarily be required. Instead, contact with another region or subunit of RNAP might function indirectly to stabilize the interaction between o region 4 and the —35 element. Several o7°-dependent activators have been shown to affect the rate of isomerization (5, 20, 30, 31). A particularly well-characterized example is provided by CRP, which possesses at least two distinct activating regions (AR1 and AR2). When bound at a so-called class II promoter, which bears a CRP recognition site that overlaps the promoter —35 region, CRP uses ARI to contact the aCTD and AR2 to contact the aNTD, the former contact mediating an effect on Kg and the latter an effect on ky (31). Moreover, it has been shown that AR2 of CRP participates in an energetically favorable interaction with RNAP (31). Activators of another class have been shown to exert their effects on the isomerization step; these activators work on promoters recognized by the alternative o factor, o>4, which appears to be unrelated to the o” class of o factors (27). Unlike PNAS | November 21,2000 | vol.97 | no.24 | 13219 GENETICS most s70-dependent activators, the s54-dependent activators generally bind well upstream of their target promoters and interact with RNAP with concomitant formation of a DNA loop (see ref. 32); furthermore, ATP hydrolysis is required for this activation. Although the mechanism of action of the s54- dependent activators is likely to be complex, it is possible that stabilization of appropriate s54–DNA contacts is a component of the activation process. It should be noted, however, that any such stabilization evidently does not occur by a tethering mechanism because s54-dependent activators can, when present at high concentrations, work directly from solution (33). In eukaryotes, as well, transcriptional activators have been implicated in post- binding steps in the initiation process (34–37), and our findings could be relevant to the understanding of how eukaryotic activators can exert these effects. A General Assay for the Interaction of DNA-Bound Regulators with s Factors. Many prokaryotic activators that bind to the DNA upstream of the promoters they regulate have binding sites that are centered roughly 40 bp upstream from the transcription start site, and some of these have been shown genetically to interact with region 4 of s70 (38–43). Our in vivo cooperative binding assay should be useful in determining whether any of these activators can stabilize the binding of region 4 to a 235 element. For those that can, our assay should also facilitate the genetic dissection of the protein–protein interaction between the acti- vator and the tethered s moiety. A potential benefit of our system is that it permits the isolation of amino acid substitutions in the s moiety of an inessential a-s chimera, the transcriptional effects of which should be limited to the test promoter. Thus, mutations affecting an essential s factor that might otherwise be pleiotropic or even lethal can be isolated and studied. We have used our experimental system to detect interactions of both s70 and s38. Interestingly, the a-s38 chimera, unlike the a-s70 chimera, activated transcription on its own from our artificial test promoter (i.e., in the absence of an adjacently bound lcI molecule). The reason for this difference is, as yet, unknown. We note that this activation-based assay may provide an especially convenient system for carrying out a genetic analysis of the protein–DNA interaction between region 4 of s38 and the 235 element. Since our in vivo assays were performed with cells that contain both plasmid-encoded a-s chimera and chromosomally encoded wild-type a, we do not know whether the activation mediated by the a-s38 chimera in the absence of lcI results from homodimeric or heterodimeric RNAP com- plexes. One s38 moiety is evidently bound to the auxiliary 235 element, and either a second s38 moiety or the aCTD may be bound nonspecifically to the DNA upstream of this 235 element (within the l operator). This hypothesis could account for the inhibitory effect of wild-type lcI on a-s38-dependent activation, as lcI might displace either a second, nonspecifically bound s38 moiety or the aCTD from the DNA, thereby reducing the magnitude of the activation. Further experiments will be re- quired to distinguish between these and other possibilities. Conclusions In summary, our findings with the a-s70 chimera define a minimal target (86 amino acids) of s70 that can interact with the activating region of lcI. Furthermore, we have shown that lcI can stabilize the binding of this region of s70 to a 235 element. We propose that the ability of lcI to stabilize the binding of region 4 of s to a 235 element can account for its stimulatory effect on transcription from PRM, and we discuss a model that can reconcile this finding with the apparently paradoxical ob- servation that lcI does not stabilize the initial binding of RNAP to PRM but rather stimulates the isomerization step. Finally, our findings validate the use of a novel genetic system that should facilitate the detection and analysis of interactions between other transcriptional regulatory proteins and various s factors from E. coli or other bacteria. We thank M. Ptashne and A. Gann for discussion, R. Kolter for plasmid pDEB2 encoding s38, and Yan Ye Xia for excellent technical assistance. This work was supported by National Institutes of Health Grant GM44025, an established investigatorship from the American Heart Association (to A.H.), and a Charles A. King Trust postdoctoral fellowship (to S.L.D.). 1. Busby, S. & Ebright, R. H. (1994) Cell 79, 743–746. 2. Ptashne, M. & Gann, A. (1997) Nature (London) 386, 569–577. 3. Hochschild, A. & Dove, S. L. (1998) Cell 92, 597–600. 4. McClure, W. R. (1985) Annu. Rev. Biochem. 54, 171–204. 5. Hawley, D. K & McClure, W. R. (1982) J. Mol. Biol. 157, 493–525. 6. Malan, T. P., Kolb, A., Buc, H. & McClure, W. R. (1984) J. Mol. Biol. 180, 881–909. 7. Busby, S. & Ebright, R. H. (1999) J. Mol. Biol. 293, 199–213. 8. Blatter, E., Ross, W., Tang, H., Gourse, R. L. & Ebright, R. H. (1994) Cell 78, 889–896. 9. Negishi, T., Fujita, N. & Ishihama, A. (1995) J. Mol. Biol. 248, 723–728. 10. Jeon, Y. H., Yamazaki, T., Otomo, T., Ishihama, A. & Kyogoku, Y. (1997) J. Mol. Biol. 267, 953–962. 11. Ross, W., Gosink, K. K., Salomon, J., Igarashi, K., Zou, C., Ishihama, A., Severinov, K. & Gourse, R. L. (1993) Science 262, 1407–1413. 12. Hochschild, A. (1994) Curr. Biol. 4, 440–442. 13. Ptashne, M. (1992) A Genetic Switch: Phage l and Higher Organisms, 2nd Ed. (Blackwell, Cambridge, MA). 14. Sauer, R. T., Jordan, S. R. & Pabo, C. O. (1990) Adv. Protein Chem. 40, 1–61. 15. Guarente, L., Nye, J. S., Hochschild, A. & Ptashne, M. (1982) Proc. Natl. Acad. Sci. USA 79, 2236–2239. 16. Hochschild, A., Irwin, N. & Ptashne, M. (1983) Cell 32, 319–325. 17. Bushman, F. D., Shang, C. & Ptashne, M. (1989) Cell 58, 1163–1171. 18. Li, M., Moyle, H. & Susskind, M. M. (1994) Science 263, 75–77. 19. Kuldell, N. & Hochschild, A. (1994) J. Bacteriol. 176, 2991–2998. 20. Li, M., McClure, W. R. & Susskind, M. M. (1997) Proc. Natl. Acad. Sci. USA 94, 3691–3696. 21. Gross, C. A., Chan, C., Dombroski, A., Gruber, T., Sharp, M., Tupy, J. & Young, B. (1998) Cold Spring Harbor Symp. Quant. Biol. 63, 141–155. 22. Dove, S. L., Joung, J. K. & Hochschild, A. (1997) Nature (London) 386, 627–630. 23. Whipple, F. W. (1998) Nucleic Acids Res. 26, 3700–3706. 24. Dove, S. L. & Hochschild, A. (1998) Genes & Dev. 12, 745–754. 25. Farrell, S., Simkovich, N., Wu, Y., Barberis, A. & Ptashne, M. (1996) Genes & Dev. 10, 2359–2367. 26. Gardella, T., Moyle, H. & Susskind, M. M. (1989) J. Mol. Biol. 206, 579–590. 27. Lonetto, M., Gribskov, M. & Gross, C. A. (1992) J. Bacteriol. 174, 3843–3849. 28. Nguyen, L. H. & Burgess, R. R. (1997) Biochemistry 36, 1748–1754. 29. Roy, S., Garges, S. & Adhya, S. (1998) J. Biol. Chem. 273, 14059–14062. 30. Shih, M-C. & Gussin, G. N. (1984) J. Mol. Biol. 172, 489–506. 31. Niu, W., Kim, Y., Tau, G., Heyduk, T. & Ebright, R. H. (1996) Cell 87, 1123–1134. 32. Rombel, I., North, A., Hwang, I., Wyman, C. & Kustu, S. (1998) Cold Spring Harb. Symp. Quant. Biol. 63, 157–166. 33. North, A. K. & Kustu, S. (1997) J. Mol. Biol. 267, 17–36. 34. Kingston, R. E. & Green, M. R. (1994) Curr. Biol. 4, 325–332. 35. Chi, T. & Carey, M. (1996) Genes & Dev. 10, 2540–2550. 36. Holstege, F. C., Fiedler, U. & Timmers, H. T. (1997) EMBO J. 16, 7468–7480. 37. Kassavetis, G. A., Kumar, A., Letts, G. A. & Geiduschek, E. P. (1998) Proc. Natl. Acad. Sci. USA 95, 9196–9201. 38. Artsimovitch, I., Murakami, K., Ishihama, A. and Howe, M. M. (1996) J. Biol. Chem. 271, 32343–32348. 39. Lonetto, M. A., Rhodius, V., Lamberg, K., Kiley, P., Busby, S. & Gross, C. (1998) J. Mol. Biol. 284, 1353–1365. 40. Landini, P. & Busby, S. J. (1999) J. Bacteriol. 181, 1524–1529. 41. Rhodius, V. A. & Busby, S. J. (2000) J. Mol. Biol. 299, 311–324. 42. Kim, S. K., Makino, K., Amemura, M., Nakata, A. & Shinagawa, H. (1995) Mol. Gen. Genet. 248, 1–8. 43. Hu, J. C. & Gross, C. A. (1985) Mol. Gen. Genet. 199, 7–13. 13220 u www.pnas.org Dove et al. most o7°-dependent activators, the o>4-dependent activators generally bind well upstream of their target promoters and interact with RNAP with concomitant formation of a DNA loop (see ref. 32); furthermore, ATP hydrolysis is required for this activation. Although the mechanism of action of the o*- dependent activators is likely to be complex, it is possible that stabilization of appropriate o**-DNA contacts is a component of the activation process. It should be noted, however, that any such stabilization evidently does not occur by a tethering mechanism because o>4-dependent activators can, when present at high concentrations, work directly from solution (33). In eukaryotes, as well, transcriptional activators have been implicated in post- binding steps in the initiation process (34-37), and our findings could be relevant to the understanding of how eukaryotic activators can exert these effects. A General Assay for the Interaction of DNA-Bound Regulators with o Factors. Many prokaryotic activators that bind to the DNA upstream of the promoters they regulate have binding sites that are centered roughly 40 bp upstream from the transcription start site, and some of these have been shown genetically to interact with region 4 of 07° (38-43). Our in vivo cooperative binding assay should be useful in determining whether any of these activators can stabilize the binding of region 4 to a —35 element. For those that can, our assay should also facilitate the genetic dissection of the protein-protein interaction between the acti- vator and the tethered o moiety. A potential benefit of our system is that it permits the isolation of amino acid substitutions in the o moiety of an inessential a-o chimera, the transcriptional effects of which should be limited to the test promoter. Thus, mutations affecting an essential o factor that might otherwise be pleiotropic or even lethal can be isolated and studied. We have used our experimental system to detect interactions of both a” and o*®. Interestingly, the a-o** chimera, unlike the a-o”° chimera, activated transcription on its own from our artificial test promoter (i.e., in the absence of an adjacently bound AcI molecule). The reason for this difference is, as yet, unknown. We note that this activation-based assay may provide . Busby, S. & Ebright, R. H. (1994) Cell 79, 743-746. . Ptashne, M. & Gann, A. (1997) Nature (London) 386, 569-577. . Hochschild, A. & Dove, S. L. (1998) Cell 92, 597-600. . McClure, W. R. (1985) Annu. Rev. Biochem. 54, 171-204. . Hawley, D. K & McClure, W. R. (1982) J. Mol. Biol. 157, 493-525. . Malan, T. P., Kolb, A., Buc, H. & McClure, W. R. (1984) J. Mol. Biol. 180, 881-909. 7. Busby, S. & Ebright, R. H. (1999) J. Mol. Biol. 293, 199-213. 8. Blatter, E., Ross, W., Tang, H., Gourse, R. L. & Ebright, R. H. (1994) Cell 78, 889-896. 9. Negishi, T., Fujita, N. & Ishihama, A. (1995) J. Mol. Biol. 248, 723-728. 10. Jeon, Y. H., Yamazaki, T., Otomo, T., Ishihama, A. & Kyogoku, Y. (1997) J. Mol. Biol. 267, 953-962. 11. Ross, W., Gosink, K. K., Salomon, J., Igarashi, K., Zou, C., Ishihama, A., Severinov, K. & Gourse, R. L. (1993) Science 262, 1407-1413. 12. Hochschild, A. (1994) Curr. Biol. 4, 440-442. 13. Ptashne, M. (1992).A Genetic Switch: Phage X and Higher Organisms, 2nd Ed. (Blackwell, Cambridge, MA). 14. Sauer, R. T., Jordan, S. R. & Pabo, C. O. (1990) Adv. Protein Chem. 40, 1-61. 15. Guarente, L., Nye, J. S., Hochschild, A. & Ptashne, M. (1982) Proc. Natl. Acad. Sci. USA 79, 2236-2239. 16. Hochschild, A., Irwin, N. & Ptashne, M. (1983) Cell 32, 319-325. 17. Bushman, F. D., Shang, C. & Ptashne, M. (1989) Cell 58, 1163-1171. 18. Li, M., Moyle, H. & Susskind, M. M. (1994) Science 263, 75-77. 19. Kuldell, N. & Hochschild, A. (1994) J. Bacteriol. 176, 2991-2998. 20. Li, M., McClure, W. R. & Susskind, M. M. (1997) Proc. Natl. Acad. Sci. USA 94, 3691-3696. 21. Gross, C. A., Chan, C., Dombroski, A., Gruber, T., Sharp, M., Tupy, J. & Young, B. (1998) Cold Spring Harbor Symp. Quant. Biol. 63, 141-155. ANnFWNE 13220 | www.pnas.org an especially convenient system for carrying out a genetic analysis of the protein-DNA interaction between region 4 of o°® and the —35 element. Since our in vivo assays were performed with cells that contain both plasmid-encoded a-o chimera and chromosomally encoded wild-type a, we do not know whether the activation mediated by the a-o*® chimera in the absence of AcI results from homodimeric or heterodimeric RNAP com- plexes. One o** moiety is evidently bound to the auxiliary —35 element, and either a second o*® moiety or the aCTD may be bound nonspecifically to the DNA upstream of this —35 element (within the A operator). This hypothesis could account for the inhibitory effect of wild-type AcI on a-o*8-dependent activation, as AcI might displace either a second, nonspecifically bound 08 moiety or the aCTD from the DNA, thereby reducing the magnitude of the activation. Further experiments will be re- quired to distinguish between these and other possibilities. Conclusions In summary, our findings with the a-o7° chimera define a minimal target (86 amino acids) of o”° that can interact with the activating region of AcI. Furthermore, we have shown that AcI can stabilize the binding of this region of a”° to a —35 element. We propose that the ability of AcI to stabilize the binding of region 4 of o to a —35 element can account for its stimulatory effect on transcription from Prm, and we discuss a model that can reconcile this finding with the apparently paradoxical ob- servation that AcI does not stabilize the initial binding of RNAP to Pam but rather stimulates the isomerization step. Finally, our findings validate the use of a novel genetic system that should facilitate the detection and analysis of interactions between other transcriptional regulatory proteins and various o factors from E. coli or other bacteria. We thank M. Ptashne and A. Gann for discussion, R. Kolter for plasmid pDEB2 encoding o°8, and Yan Ye Xia for excellent technical assistance. This work was supported by National Institutes of Health Grant GM44025, an established investigatorship from the American Heart Association (to A.H.), and a Charles A. King Trust postdoctoral fellowship (to S.L.D.). 22. Dove, S. L., Joung, J. K. & Hochschild, A. (1997) Nature (London) 386, 627-630. 23. Whipple, F. W. (1998) Nucleic Acids Res. 26, 3700-3706. 24. Dove, S. L. & Hochschild, A. (1998) Genes & Dev. 12, 745-754. 25. Farrell, S., Simkovich, N., Wu, Y., Barberis, A. & Ptashne, M. (1996) Genes & Dev. 10, 2359-2367. 26. Gardella, T., Moyle, H. & Susskind, M. M. (1989) J. Mol. Biol. 206, 579-590. 27. Lonetto, M., Gribskov, M. & Gross, C. A. (1992) J. Bacteriol. 174, 3843-3849. 28. Nguyen, L. H. & Burgess, R. R. (1997) Biochemistry 36, 1748-1754. 29. Roy, S., Garges, S. & Adhya, S. (1998) J. Biol. Chem. 273, 14059-14062. 30. Shih, M-C. & Gussin, G. N. (1984) J. Mol. Biol. 172, 489-506. 31. Niu, W., Kim, Y., Tau, G., Heyduk, T. & Ebright, R. H. (1996) Cell 87, 1123-1134. 32. Rombel, I., North, A., Hwang, I., Wyman, C. & Kustu, S. (1998) Cold Spring Harb. Symp. Quant. Biol. 63, 157-166. 33. North, A. K. & Kustu, S. (1997) J. Mol. Biol. 267, 17-36. 34. Kingston, R. E. & Green, M. R. (1994) Curr. Biol. 4, 325-332. 35. Chi, T. & Carey, M. (1996) Genes & Dev. 10, 2540-2550. 36. Holstege, F. C., Fiedler, U. & Timmers, H. T. (1997) EMBO J. 16, 7468-7480. 37. Kassavetis, G. A., Kumar, A., Letts, G. A. & Geiduschek, E. P. (1998) Proc. Natl. Acad. Sci. USA 95, 9196-9201. 38. Artsimovitch, I., Murakami, K., Ishihama, A. and Howe, M. M. (1996) J. Biol. Chem. 271, 32343-32348. 39. Lonetto, M. A., Rhodius, V., Lamberg, K., Kiley, P., Busby, S. & Gross, C. (1998) J. Mol. Biol. 284, 1353-1365. 40. Landini, P. & Busby, S. J. (1999) J. Bacteriol. 181, 1524-1529. 41. Rhodius, V. A. & Busby, S. J. (2000) J. Mol. Biol. 299, 311-324. 42. Kim, S. K., Makino, K., Amemura, M., Nakata, A. & Shinagawa, H. (1995) Mol. Gen. Genet. 248, 1-8. 43. Hu, J. C. & Gross, C. A. (1985) Mol. Gen. Genet. 199, 7-13. Dove et al.