Changing the mechanism of transcriptional activation by phage λ repressor

Abstract

The first steps of transcription initiation include binding of RNA polymerase to a promoter to form an inactive, unstable, closed complex (described by an equilibrium constant, K_B) and isomerization of the closed complex to an active, stable, open complex (described by a forward rate constant, k_f). λ cI protein activates the P_RM promoter by specifically increasing k_f. A positive control mutant, cI-pc2, is defective for activation because it fails to raise k_f. An Arg to His change in the σ⁷⁰ subunit of RNA polymerase was previously obtained as an allele-specific suppressor of cI-pc2. To elucidate how the mutant polymerase restores the activation function of the mutant activator, abortive initiation assays were performed, using purified cI proteins and RNA polymerase holoenzymes. The change in σ does not significantly alter K_B or k_f in the absence of cI protein. As expected, cI-pc2 activates the mutant polymerase in the same way that wild-type cI activates the wild-type polymerase, by increasing k_f. An unexpected and novel finding is that the wild-type activator stimulates the mutant polymerase, but not wild-type polymerase, by increasing K_B.

Gene expression is frequently regulated by activator proteins that bind to specific DNA sites and increase the rate of transcription initiation by RNA polymerase at a nearby promoter. In most cases, activation is thought to involve a favorable protein–protein interaction between the activator and RNA polymerase on the template DNA. A classic example is phage λ cI protein (also called λ repressor), which activates transcription of its own gene from P_RM, the promoter for synthesis of repressor during maintenance of lysogeny (1). P_RM is a weak, leftward promoter that is oriented divergently from a strong, rightward promoter (P_R) in the O_R (right operator) region (Fig. 1). Both promoters are controlled by binding of cI protein to three related 17-bp sites. P_R overlaps with O_R1 and O_R2, while P_RM overlaps with O_R2 and O_R3. At the normal concentration of cI in a lysogen, cI dimers bind cooperatively to O_R1 and O_R2, leaving O_R3 vacant most of the time. P_R is severely repressed, while the cI dimer bound to O_R2 activates P_RM. At higher concentrations, cI also binds to O_R3, thereby repressing P_RM.

Figure 1.

Sequence of the λ O_R region. The 5′ ends of RNAs made by natural transcription are shown, as well as trinucleotides made in abortive initiation reactions. Critical promoter elements are marked with asterisks. The O_R3-r2 mutation was used in assays in vitro.

Activation of P_RM involves an interaction between an acidic patch on the surface of the helix–turn–helix motif of cI protein (2–4) and a target region near the C terminus of the σ⁷⁰ subunit of Escherichia coli RNA polymerase (5, 6). We previously showed that a single amino acid change in σ (Arg-596 to His; abbreviated RH596) restores activation by a mutant form of cI that has a single amino acid change in the activation patch (Asp-38 to Asn; called pc2 for its specific defect in positive control). The change in σ fully suppressed the cI-pc2 defect in vivo, allowing the mutant activator to activate a P_RM lacZ fusion. Two other mutant activators, cI-pc1 (2) and cI-pc3 (3), were not suppressed by the mutant σ. In this paper, we address two questions that could not be resolved by genetic experiments. The first issue was the formal possibility that the mutant σ might exert its effect indirectly (e.g., by altering expression of other genes, since nearly all transcription is σ⁷⁰-dependent). This idea is ruled out here by the demonstration that the mutant activator stimulates initiation by the mutant polymerase in a purified system in vitro. The second question was whether cI-pc2 stimulates the mutant polymerase by the same mechanism used by wild-type cI to stimulate the wild-type polymerase.

λ cI activates P_RM by speeding up the rate at which RNA polymerase forms an “open,” transcriptionally active enzyme–DNA complex. According to a simplified two-step model for open complex formation, RNA polymerase (R) binds to the promoter (P) to form an inactive, unstable closed complex, RP_C, that then undergoes a relatively slow isomerization to form the open complex, RP_O:

Initial binding is characterized by an apparent equilibrium constant, K_B; isomerization is characterized by a forward rate constant, k_f. When wild-type E. coli RNA polymerase was used, Hawley and McClure (7) found that wild-type cI activates P_RM by enhancing k_f about 10-fold, without having a significant effect on K_B. The cI-pc2 mutant fails to stimulate k_f without creating a severe defect in K_B (8). In the experiments reported here, similar results were obtained using wild-type RNA polymerase. In addition, we show that (i) the Arg-596 to His change in σ does not significantly alter P_RM promoter strength in the absence of cI protein; and (ii) with the mutant polymerase, cI-pc2 raises k_f without changing K_B significantly, which implies that cI-pc2 stimulates the mutant enzyme by the same mechanism that wild-type cI stimulates the wild-type enzyme. An interesting and unexpected discovery is that wild-type cI increases K_B for the mutant polymerase, instead of activating the isomerization step.

MATERIALS AND METHODS

Reagents.

TE is 10 mM Tris·HCl, pH 8.0/0.1 mM EDTA. Lysis buffer is 100 mM Tris·HCl, pH 8.0/200 mM NaCl/10 mM MgCl₂/2 mM CaCl₂/1 mM EDTA/0.1 mM dithiothreitol/glycerol (5%, vol/vol)/hen egg white lysozyme (150 μg/ml). SB is 10 mM Tris·HCl, pH 8.0/2 mM CaCl₂/0.1 mM EDTA/0.1 mM dithiothreitol/5% glycerol. Standard reaction buffer is 40 mM Hepes, pH 8.0/100 mM potassium glutamate/10 mM MgCl₂/1 mM dithiothreitol/bovine serum albumin (30 μg/ml). Cytidylyl(3′-5′)adenosine (CpA) and uridylyl(3′-5′)adenosine (UpA) were purchased from ICN and Sigma, respectively.

Plasmids and Bacteria.

Plasmids are listed in Table 1. A pair of isogenic E. coli strains, rpoD⁺ and rpoD-RH596, were constructed by phage P1 transduction, using zgh::Tn10, an insertion that is linked to rpoD and confers resistance to tetracycline (20 μg/ml). Donors were CAG4018 (zgh::Tn10) and CAG4025 (zgh::Tn10 rpoD2, which is identical to rpoD-RH596) (15) carrying pMS1382, an rpoD⁺ plasmid that suppressed the temperature-sensitive phenotype of the rpoD2 strain and enhanced growth of P1. The recipient was E. coli RV308, which is Δ(lac)χ74 galPO-308::IS2 rpsL (16). Tetracycline-resistant transductants were checked for absence of pMS1382 and P1, and rpoD genotypes were determined by PCR amplification of chromosomal DNA and PCR sequencing, using the Promega fmol system. The resulting strains are MS4274 (RV308 zgh::Tn10) and MS4276 (RV308 zgh::Tn10 rpoD2). Although the rpoD2 (RH596) mutation does not confer a temperature-sensitive phenotype in the RV308 genetic background, both strains were maintained at 30°C.

Table 1.

Plasmids

Plasmid	Description
pAED4	pBR322-based Ampr T7 expression vector (D. S. Doering); derivative of pET-3a (9)
pCICK	pACYC177-based Camr plasmid carrying λ cI-ts857 (10)
pLysS	pACYC184-based Camr plasmid carrying T7 gene 3.5 (11)
pMRG8	pBR322-based Ampr plasmid carrying E. coli rpoD+ fused to λ PLOL (12)
pMS802	pSC101-based Spcr plasmid carrying E. coli lacIq (13)
pMS1382	pBR322-based Ampr plasmid carrying E. coli rpoD+ fused to PlacUV5
pMS1454	NdeI–KpnI cI fragment* in the NdeI–KpnI backbone of pAED4
pMS1456	NdeI–KpnI cI-pc2 fragment* in the NdeI–KpnI backbone of pAED4
pMS1488	Replace BamHI–HindIII rpoD+ fragment of pMRG8 with the corresponding BamHI–BssHII rpoD-RH596 fragment from pMS1366 (5). HindIII and BssHII ends were filled in, using DNA polymerase I large fragment.
pPR1347	pSC101-based Kanrcosλ cosmid carrying Salmonella typhimurium rfb and rfc operons (14)

Antibiotic resistance: Amp, ampicillin; Cam, chloramphenicol; Spc, spectinomycin; and Kan, kanamycin. 

^*Fragment extends from a site-directed NdeI site overlapping the start codon to a KpnI linker inserted at a filled-in MunI site downstream of cI. The gene contains four site-directed mutations that create restriction sites but do not alter the amino acid sequence. 

These E. coli strains were made transiently sensitive to Salmonella phage P22, using cosmid pPR1347 (14), which directs synthesis of the O antigen of S. typhimurium. The cosmid is maintained by selecting resistance to kanamycin (10 μg/ml), but is spontaneously lost when kanamycin is removed. MS4274 and MS4276 were transduced to kanamycin resistance with λ particles containing pPR1347. The resulting strains were purified, checked for absence of λ, and infected in liquid culture with high multiplicities of one or both of the following: P22 P_RM-lac8 ΔAB (5), which carries E. coli lacZ fused to λ P_RMO_R; and P22 P_lac-cIA (5), which carries λ cI fused to a weak constitutive version of the E. coli lac promoter, and also carries the bla gene for resistance to ampicillin. Infected cultures were diluted, incubated overnight at 30°C, and plated on LB agar (17) containing 5-bromo-4-chloro-3-indolyl β-d-galactopyranoside (150 μg/ml; to detect presence of the lacZ prophage) with or without ampicillin (50 μg/ml; to select for the second prophage). Candidates for double lysogens were purified repeatedly on LB agar and screened for retention of both prophages by testing for retention of lacZ and bla; unstable double lysogens (which probably carried tandem prophages that were subsequently resolved by recombination) were discarded. Stable double lysogens presumably have nontandem prophages: P_RM-lac8 ΔAB at the P22 attachment site and P_lac-cIA at the λ attachment site (Fig. 2). [A recombinant prophage containing both fusions cannot arise, because P_RM-lac8 ΔAB att²² is allelic to P_lac-cIA att^λ (5).]

Figure 2.

System used to measure activation in vivo. Strains have only one form of σ⁷⁰, either wild-type or mutant, encoded by the rpoD gene at its normal locus in the E. coli chromosome (thick line). A prophage (thin line) at the λ attachment site produces a low level of cI protein encoded by various cI alleles fused to a weak, constitutive promoter. Effective combinations of σ and cI activate transcription of a lacZ reporter gene fused to wild-type λ P_RMO_R on a prophage at the P22 attachment site. λ cI protein does not control its own synthesis and does not control maintenance of lysogeny by either prophage.

Overexpression plasmids pMS1454 and pMS1456 (which carry λ cI fused to the T7 φ10 promoter and gene 10 ribosome binding site) were used in derivatives of E. coli BL21(λDE3)/pLysS (11) that also carry pMS802, a compatible lacI^q plasmid that improved retention of pMS1454 and pMS1456. These strains, MS4456 and MS4458, were propagated in media containing chloramphenicol (20 μg/ml), spectinomycin (50 μg/ml), and ampicillin and were used to overproduce wild-type cI and cI-pc2 protein, respectively. [λDE3 does not carry the λ cI gene (9).]

Overexpression plasmids pMRG8 and pMS1488 carry E. coli rpoD⁺ and rpoD-RH596, respectively, fused to λ P_L, which is controlled by temperature-sensitive λ repressor encoded by pCICK. E. coli MS4522 is MS4274/pCICK/pMRG8; MS4523 is MS4276/pCICK/pMS1488. Because the rpoD allele in the E. coli chromosome is the same as the rpoD allele in the overproducing plasmid, each strain produces only one type of σ⁷⁰, either wild-type or mutant.

RNA Polymerase Holoenzymes.

σs were overproduced in parallel in MS4522 and MS4523 and were purified as described by Gribskov and Burgess (12) through the DEAE-Sephadex column step, with minor modifications. Partially purified σs were mixed with the same batch of purified wild-type E. coli RNA polymerase core enzyme (18) at a molar ratio of 1.5. The resulting holoenzymes were then purified (18). Each was estimated to be >90% pure as judged by Coomassie brilliant blue staining of samples analyzed by SDS/PAGE. Fractional activity of the holoenzymes, assayed on a poly[d(AT)] template (19), was about 40%. The active concentrations were used for the TAU plot analyses that yielded the reported values for K_B.

Purification of cI.

MS4456 and MS4458 were grown at 37°C in LB plus antibiotics (see above) until OD₅₅₀ reached 0.7, induced by adding isopropyl β-d-thiogalactopyranoside to 0.5 mM, incubated for 3 hr, and harvested by centrifugation. Methods for cell lysis and purification of cI were adapted from ref. 20. Cells were suspended in lysis buffer, sodium deoxycholate (0.05%) and phenylmethylsulfonyl fluoride (35 μg/ml) were added, and mixtures were incubated at 4°C for 20 min. To reduce viscosity, crude lysates were sonicated, diluted ≈5-fold with SB + 0.2 M NaCl, and treated with DNase I (50 μg/ml) at 4°C for 30 min. NaCl concentration was adjusted to 0.55 M before low-speed centrifugation to remove debris. Cleared lysates were mixed with polyethyleneimine (0.1%) at 4°C and the resulting precipitates were removed by centrifugation. Ammonium sulfate was added (40 g per 100 ml of supernatant), and the resulting precipitates were collected by centrifugation, dissolved in SB + 0.2 M NaCl, and dialyzed against SB + 0.1 M NaCl overnight at 4°C. Dialysates were clarified by centrifugation and loaded on Affi-Gel Blue (Bio-Rad) columns equilibrated with SB + 0.1 M KCl. Columns were washed with 2 bed vol of SB + 0.1 M KCl and eluted with SB containing a linear gradient of KCl (0.1 to 1 M). Fractions containing cI (identified by SDS/PAGE) were pooled, diluted 2-fold with SB, and loaded on HA-Ultrogel (Sigma) columns equilibrated with SB + 0.1 M KCl. Columns were washed with 1 bed vol of 0.1 M potassium phosphate, pH 7.4, and eluted with a linear gradient of potassium phosphate (0.1 to 1 M), pH 7.4. Peak fractions containing cI were pooled and concentrated by precipitation with ammonium sulfate. Pellets were dissolved in SB + 0.2 M NaCl (≈4 mg of protein per ml) and dialyzed against SB + 0.2 M NaCl overnight at 4°C. Dialysates were clarified by centrifugation and supernatants were stored frozen at −20°C (short-term) or −70°C (long-term). Both cI preparations were estimated to be about 90% pure. Activities were determined by repression assays, which correlate well with filter-binding assays of DNA-binding activity (8): 11 nM cI gives 50% repression of P_R, using wild-type RNA polymerase (see Fig. 3A). The active fraction (measured by comparing activity with A₂₈₀) was 15% for wild-type cI and 30% for cI-pc2.

Figure 3.

Effect of wild-type cI and cI-pc2 on repression of P_R and activation of P_RM. The activities of P_R (A and C) and P_RM (B and D) in the presence of various concentrations of wild-type cI (•) or cI-pc2 (○) were measured by monitoring the rate of synthesis of abortive products by wild-type RNA polymerase holoenzyme (A and B) or holoenzyme containing σ⁷⁰-RH596 (C and D) as described in the text. The maximal rate for P_R (set at 100) corresponds to 361 or 337 CpApU per promoter per min in A and C, respectively; the nonactivated rate for P_RM (set at 1) corresponds to 71 or 81 UpApU per promoter per min in B and D, respectively.

DNA Template.

A fragment containing λ P_RM and P_R was obtained by PCR amplification, using a HindIII digest of λ112 cI-sus34 O_R3-r2 (16) DNA as template (7.9 pM) and limiting amounts of primers (200 nM each). The product (343 bp; +140 to −203 with respect to P_RM) was extracted with phenol/chloroform (1:1) and chloroform, and dialyzed against TE + 0.1 M NaCl and TE. DNA concentration was determined by measuring A₂₆₀ and confirmed by gel electrophoresis.

Activation-Repression Curves.

Effects of cI on P_RM and P_R were measured by monitoring the rate of synthesis of abortive products as described (7, 8). DNA, UpA or CpA, [α-³²P]UTP, and cI were mixed and prewarmed at 37°C for 8 min in 40 μl of standard reaction buffer. Reactions were initiated by adding 10 μl of RNA polymerase prewarmed in standard reaction buffer. Final concentrations were 1.1 nM DNA, 0.5 mM UpA or CpA, 50 μM [α-³²P]UTP (≈400 cpm/pmol), and 40 nM RNA polymerase. The P_RM reaction (UpA + UTP → UpApU) was assayed after 6 and 12 min. The P_R reaction (CpA + UTP → CpApU) was assayed after 8 and 16 min. Product trinucleotides were separated from UTP by ascending chromatography (8), and radioactivity in the product peak was normalized to total radioactivity on the chromatogram (≈300,000 cpm). Reaction rates (product per promoter per min) were calculated, normalized to the controls (without added cI), and plotted against the concentration of active cI.

TAU Plot Analysis.

The average time required for open complex formation (τ_OBS) at P_RM was measured by lag assays (7). DNA, nucleotides, and cI were mixed in 256 μl of standard reaction buffer and prewarmed at 37°C for 8 min before initiating the reaction by addition of 64 μl of RNA polymerase prewarmed in standard reaction buffer. Final concentrations were 1 nM DNA, 0.2 or 0.5 mM UpA, 50 μM [α-³²P]UTP (≈400 cpm/pmol), and 0 or 50 nM cI protein; RNA polymerase was varied from 12 to 80 nM. UpA concentration was 0.2 mM in reactions without cI and 0.5 mM in reactions with cI; in both cases, substrate depletion was <12%. Portions (20 μl) were removed at various times and spotted onto the origin of a chromatogram prestreaked with 20 μl of 0.1 M EDTA to stop the reaction. The percentage of UTP incorporated (see above) was plotted versus time for each reaction, and τ_OBS was determined by a computer program (7) that performed a least-squares fit of the data to the equation

where N is the amount of UpApU synthesized per promoter, V is the final steady-state velocity, and t is time (min). Values of τ_OBS were measured at various RNA polymerase concentrations ([R]) and were plotted against 1/[R]. These TAU plots produce straight lines, as predicted by the equation

The reciprocal of the intercept yields k_f and the ratio of intercept to slope yields K_B. In this study, τ_OBS values were uniformly shorter and k_f values were uniformly higher than those obtained previously (7, 8). Our standard reaction buffer contained potassium glutamate instead of KCl, to minimize experimental variability due to small differences in chloride concentrations (21); this change probably contributes to the increase in k_f (unpublished data). Another difference is that the template used here has the O_R3-r2 mutation, which was not present in any of the templates used previously in abortive initiation assays (7, 8). The O_R3 mutation does not block activation of P_RM by wild-type cI, but reduces binding of cI to O_R3 and repression of P_RM (16). Though O_R3-r2 is at a nonconserved position (−18; Fig. 1), it may slightly improve P_RM activity (16).

RESULTS

Effect of σ⁷⁰-RH596 in E. coli.

In the experiments reported previously, the effect of the Arg-596 to His change in E. coli σ⁷⁰ was examined in S. typhimurium strains carrying the E. coli rpoD gene on a high-copy plasmid and the S. typhimurium rpoD⁺ gene in the chromosome (5). To examine the effect of the mutant σ in E. coli strains that have only one copy of rpoD, we made use of the fact that the rpoD-RH596 mutation was previously isolated as a viable, haploid allele; the rpoD2 mutation characterized by Hu and Gross (15) is identical to rpoD-RH596. Isogenic E. coli rpoD⁺ and rpoD2 strains were constructed in a Δ(lac) genetic background as described above. These strains were then lysogenized with phages that provide a system for assaying activation of P_RM by λ cI in vivo (Fig. 2).

The results obtained using E. coli haploids (Table 2) are strikingly similar to those obtained using rpoD plasmids in S. typhimurium (5). σ⁷⁰-RH596 fully restores the activation function of cI-pc2. This effect is specific: the mutant σ does not respond to cI-pc1 or -pc3 and is slightly defective in responding to wild-type cI. The basal activity of P_RM (activity with no cI) is not affected by the change in σ.

Table 2.

Effect of rpoD-RH596 in E. coli haploids

Activator	Relative β-galactosidase activity
	rpoD+	rpoD-RH596
None	15.0  ±  1.1	15.6  ±  1.6
cI wild type	(100)	79.1  ±  4.3
cI-pc1 (Gly-43 to Arg)	4.7  ±  1.1	5.4  ±  1.2
cI-pc2 (Asp-38 to Asn)	15.5  ±  0.7	133  ±  10
cI-pc3 (Glu-34 to Lys)	0.6  ±  0.1	2.0  ±  0.5

Strains are derivatives of MS4274 and MS4276 carrying a P_RM-lac8 ΔAB prophage plus either no second prophage or one of four P_lac-cIA prophages. Four cultures of each strain were grown at 30°C in M9CAA (22), harvested in exponential phase, and assayed as described (23). Specific activities of the four control cultures (rpoD⁺ cI⁺) were averaged to yield the mean specific activity of the control (set at 100; corresponds to 131 Miller units). Specific activity of each of the other cultures was expressed as a percentage of the mean specific activity of the control. The four percentages for a particular genotype were used to calculate the mean and standard deviation given. 

We note that the cI-pc1 and -pc3 mutants interfere with P_RM activity, reducing expression below the basal level. The repression observed here, about 70% for cI-pc1 and 90% for -pc3, was comparable with both σs. This effect has been seen previously (3, 5, 24). The possibility that cI-pc1 and -pc3 inhibit P_RM, rather than simply fail to activate P_RM, led to the original choice of cI-pc2 for kinetic studies of the positive control defect (8). It has been suggested that cI-pc1 and -pc3 repress P_RM by binding to O_R3, but the evidence presented does not strongly support that model (24). An alternative model is that cI-pc1 and -pc3 bound to O_R2 (and not to O_R3) have an unfavorable interaction with RNA polymerase, because each mutation introduces a basic amino acid residue in the patch of cI that normally contacts a basic patch of RNA polymerase.

Effect of σ⁷⁰-RH596 on Activation and Repression in Vitro.

To show that σ⁷⁰-RH596 directly restores the activation function of cI-pc2, abortive initiation assays were performed, using purified cI proteins (wild-type or cI-pc2) and purified RNA polymerase holoenzymes containing σ⁷⁰ (wild-type or σ⁷⁰-RH596). On a template carrying both P_R and P_RM, open complexes at each promoter were monitored by their ability to repeatedly synthesize a trinucleotide corresponding to −1, +1, and +2: open complexes at P_R convert CpA and UTP to CpApU, whereas open complexes at P_RM convert UpA and UTP to UpApU (see Fig. 1). Fig. 3 shows the effects of the cI proteins on occupancy of both promoters. P_R was repressed by increasing amounts of either wild-type cI or cI-pc2, and the repression curves were similar when wild-type RNA polymerase (Fig. 3A) and RNA polymerase containing σ⁷⁰-RH596 (Fig. 3C) were used. With wild-type RNA polymerase, P_RM was activated 3-fold by wild-type cI, but less than 2-fold by cI-pc2 (Fig. 3B). With the mutant RNA polymerase, P_RM was activated about 5-fold by cI-pc2, but less than 2-fold by wild-type cI (Fig. 3D). These results are in excellent agreement with the activation assays in vivo (Table 2). The activation-repression experiments of Fig. 3 cannot be used to determine the full extent or identity of the activated step, but these results indicated the choice of cI concentration to be used for the kinetic experiments described next.

Kinetics of Open Complex Formation.

Lag assays were used to determine τ_OBS, the average time required for RNA polymerase to form an open complex at P_RM in the presence or absence of cI protein. τ_OBS was measured at various RNA polymerase concentrations and plotted against the reciprocal of RNA polymerase concentration. These “TAU plots” were used to calculate the apparent binding constant, K_B, and isomerization rate constant, k_f (Table 3).

Table 3.

Kinetic parameters of open complex formation at P_RM

Activator	Wild-type RNA polymerase		Mutant RNA polymerase
	KB × 10−6, M−1	kf × 103, s−1	KB × 10−6, M−1	kf × 103, s−1
None	13	5.0	7.3	5.6
cI wild type	11	86	63	13
cI-pc2	25	5.0	4.3	150

In the absence of cI, the values for K_B and k_f obtained with the mutant RNA polymerase and those obtained with wild-type polymerase are the same within experimental error (estimated as ±30–40% from the linear least-squares analyses). This finding agrees with the results of Table 2, which show that the mutant σ had no effect on the basal activity of P_RM in vivo.

With wild-type RNA polymerase, activation of P_RM occurred as described previously (7, 8). Wild-type cI increased k_f more than 10-fold, but it had no significant effect on K_B. The cI-pc2 mutant activator failed to raise k_f, but it did not create a defect in initial binding.

The mutant RNA polymerase is strikingly different from wild-type polymerase in its response to wild-type cI and cI-pc2. The Arg-596 to His change in σ fully restores the activation function of cI-pc2. With the mutant polymerase, the mutant activator raised k_f more than 20-fold and had no significant effect (within experimental error) on K_B. Thus, cI-pc2 stimulates open complex formation at P_RM by the mutant polymerase in the same way that wild-type cI stimulates the wild-type polymerase. Somewhat surprisingly, in the presence of the mutant RNA polymerase, wild-type cI raised K_B about 9-fold and k_f only about 2-fold.

DISCUSSION

Activation of P_RM was investigated with all possible combinations of wild-type and mutant RNA polymerase holoenzymes and wild-type and mutant cI proteins. As shown previously, wild-type cI stimulates wild-type RNA polymerase to form open complexes on P_RM by increasing k_f (7), whereas the mutant cI-pc2 activator does not (8). With holoenzyme containing σ⁷⁰-RH596, however, cI-pc2 stimulates formation of open complexes on P_RM by increasing k_f more than 20-fold, with little or no effect on K_B (Table 3). This result shows that the change in σ directly suppresses the cI-pc2 defect, since the only proteins in the abortive initiation assays are RNA polymerase and cI. Because the same rate parameter is affected, cI-pc2 apparently stimulates the mutant polymerase in the same way that wild-type cI stimulates the wild-type polymerase. The cI-pc2 defect is fully suppressed, both in vivo and in vitro: the mutant activator works with the mutant polymerase slightly better than the wild-type activator works with the wild-type polymerase. The mutant σ has no significant effect on the basal activity of P_RM, in vivo or in vitro. Thus, σ⁷⁰-RH596 specifically, fully, and mechanistically suppresses the cI-pc2 defect.

Two general types of models can be proposed to explain the activation defect of cI-pc2 and the restoration of activation by the change in σ. The first model envisions that the pc2 mutation (Asp-38 to Asn) removes a favorable contact between the activator and RNA polymerase, and the Arg-596 to His change in σ provides an alternative, equally favorable contact. The second model is that Asp-38 of cI does not normally play a critical role in activation, but the Asp to Asn change (pc2) introduces an unfavorable interaction that prevents activation. Replacing Arg-596 of σ with His could simply relieve the clash without creating a new, favorable contact. Either model can explain the observation that activation of the mutant polymerase by the mutant activator resembles activation of the wild-type polymerase by the wild-type activator. Both models imply that the activation patch of cI is close to residue 596 of σ, an idea that is supported by molecular modeling (25). Other mutational studies (ref. 4; unpublished data) favor the second model, because activation works reasonably well with several other combinations of amino acids at position 596 of σ and position 38 of cI. Of course, elements of both models may apply here, just as they do in detailed studies of other protein–protein or protein–nucleic acid interactions.

A novel and surprising finding is that the mutant polymerase differs from the wild-type polymerase in its response to wild-type cI. With the wild-type enzyme, cI raises k_f and has little effect on K_B; with the mutant enzyme, cI raises K_B about 9-fold and has a small effect on k_f. The surprising aspect of this finding is that a single amino acid change in σ switches the activation of P_RM from the isomerization step to the initial binding step. Although similar switches have not yet been reported for other combinations of polymerases and activators, such changes in simpler enzyme–substrate interactions are well documented (26). In those cases, a structural change that favorably affects the interaction between a substrate and an enzyme subsite can decrease K_m (analogous to an increase in K_B) or increase k_cat (analogous to k_f) or both. The partitioning between the two steps depends on whether the change in structure stabilizes the ground state or the activated enzyme–substrate complex. Applying the same reasoning to interactions between cI and RNA polymerase at P_RM, cI wild-type stabilizes the transition state between RP_C and RP_O for the wild-type enzyme by about 1.6 kcal/mol; with the mutant enzyme, wild-type cI stabilizes the closed complex (RP_C) by about 1.3 kcal/mol. The observation that single amino acid changes in σ or in cI can affect the partitioning between the two steps implies that favorable interactions affecting the first step are not profoundly different from those affecting the second step.

Initial binding of polymerase to the promoter is conceptually simpler than isomerization, since initial binding presumably resembles DNA binding by simpler proteins (e.g., cI binding to its operators). Enhancement of initial polymerase–promoter binding by an activator is also relatively easy to understand. For example, a better fit between complementary patches on the activator and polymerase is readily grasped as leading to an increase in a binding constant. The real mystery in promoter function and activation is the nature of the isomerization step, which involves separation of the DNA strands around the startpoint of transcription, uptake of ions, and possibly a conformational change in the enzyme (27). These reactions are not fully understood, and neither are the mechanisms by which activators speed them up. If additional amino acid changes in activators and polymerases can be found that have discrete effects on the steps leading to the open complex, it may be possible to interpret the partitioning between the steps in terms of structural changes that go beyond the current, still rather nebulous, descriptions.