Multilevel autoregulation of λ repressor protein CI by DNA looping in vitro

Abstract

The prophage state of bacteriophage λ is extremely stable and is maintained by a highly regulated level of λ repressor protein, CI, which represses lytic functions. CI regulates its own synthesis in a lysogen by activating and repressing its promoter, P_RM. CI participates in long-range interactions involving two regions of widely separated operator sites by generating a loop in the intervening DNA. We investigated the roles of each individual site under conditions that permitted DNA loop formation by using in vitro transcription assays for the first time on supercoiled DNA that mimics in vivo situation. We confirmed that DNA loops generated by oligomerization of CI bound to its operators influence the autoactivation and autorepression of P_RM regulation. We additionally report that different configurations of DNA loops are central to this regulation—one configuration further enhances autoactivation and another is essential for autorepression of P_RM.

Bacteriophage lambda (λ) of Escherichia coli can grow either in a lytic or lysogenic mode. In a lysogenic cell, the phage-encoded λ repressor protein (CI) prevents lytic growth by directly repressing two promoters needed to express lytic functions, P_L and P_R (1–3). Each promoter is associated with a CI recognition site or operator, O_L (composed of three adjacent subsites, O_L1, O_L2, O_L3) and O_R (O_R1, O_R2, O_R3). O_R is associated with promoter P_RM (promoter for maintenance of repressor synthesis), which directs transcription of the cI gene in the prophage state (4–6). Each subsite binds a CI dimer (7).

The CI protein autoregulates its synthesis. At low cellular CI concentration, CI enhances its own synthesis from P_RM; when high, CI represses P_RM (1, 8, 9). It was originally believed that both positive and negative autoregulations are achieved exclusively by the action of CI dimers at the P_RM-O_R-P_R sequence of the phage genome (1, 10) (Fig. 1A), based on the following observations. (i) There is a hierarchy of intrinsic binding affinities of a CI dimer to individual operators sites: O_R1 > O_R2 > O_R3 (11–17); (ii) CI bound to the intrinsically weak O_R2 site is strengthened by cooperative interactions with CI bound to the stronger adjacent O_R1 site, and the ensemble of two CI dimers at the O_R1 ∼ O_R2 sites represses P_R and activates P_RM (13, 16) (Fig. 1A); (iii) at high CI concentrations, a CI dimer can bind to the weakest operator site, O_R3, repressing P_RM (1, 2) (Fig. 1C). Incidentally, a second pair of CI dimers binds cooperatively to O_L1 ∼ O_L2, and represses P_L (7).

Fig. 1.

Models of CI regulation by DNA looping. Detailed conformations of the structures are not known and maps are not drawn to scale. (A) Promoters (P_R, P_L, and P_RM); operators O_L (O_L1, O_L2, O_L3) and O_R (O_R1, O_R2, O_R3) are in blue rectangles; CI dimers (one monomer is shown in yellow, the other is in gray). The bent arrows show the transcription start points of promoters. The dashed line indicates transcripts from P_R, P_L, and P_RM. The cI gene is transcribed from P_RM. (B) DNA looping and octamer formation (8-mer) by CI tetramer binding to O_L1 ∼ O_L2 interacts with that at O_R1 ∼ O_R2. (C) Octamer and tetramer (12-mer) of CI binding to O_L and O_R. Red X means promoter is turned off. ΔG_oct and ΔG_tetr values for octamer and tetramer loops stability are from Zurla et al. (25), measured on linear DNA under different conditions. See refs. 7, 18, 20, and 59 for models on CI regulation.

This model was recently modified on the basis of physical and genetic experiments showing that a CI tetramer cooperatively bound to O_R1 ∼ O_R2 interacts with a tetramer cooperatively bound to O_L1 ∼ O_L2, located 2.3 kbp away (18–21), resulting in the looping out of the intervening DNA, as shown by electron and atomic force microscopy (18, 22, 23). DNA loop was also observed by cooperative interactions of CI at sites separated by only five and six helical turns (23, 24). The kinetic and thermodynamic properties of loop formation were determined by single DNA molecule analysis (25, 26). Mathematical modeling and genetic studies suggested that DNA looping further enhances autoactivation of P_RM (27, 28). Genetic experiments also indicated that O_R1 ∼ O_R2/O_L1 ∼ O_L2 mediated DNA looping helps CI autorepression by stabilizing CI dimer binding at O_R3 through a cooperative interaction with a CI dimer bound to O_L3 (19, 20) (Fig. 1C).

We previously investigated the thermodynamic parameters of DNA loops with or without the involvement of O_L3 and O_R3 by tethered particle microscopy of single DNA molecules (25). The results showed that, whereas the DNA loops, which involve all six operators and six CI dimers are relatively stable (Fig. 1C), those without O_R3 and O_L3 and involving only four dimers are less stable (Fig. 1B). However, the stabilities were determined on linear DNA compared to the transcription experiments (see below) performed using supercoiled DNA templates. These studies cannot provide insight into the properties of the P_RM promoter itself. The original model of P_RM regulation by CI was based on studies using DNA templates carrying the O_R region only (1–3, 7). To study the involvement of O_L in the regulation of P_RM, we reasoned that the weak interaction forming the octamer might not allow us to detect activation caused by looping using linear templates. Because it is believed that supercoiling increases the local concentration of distant sites (29), we chose to use supercoiled templates in a transcription assay to more closely reflect the in vivo situation. Thus, we developed an in vitro transcription system using supercoiled templates. With this system, we have confirmed in vitro the enhancement of P_RM autoactivation by the less stable DNA loop mediated by four CI dimers, which forms at physiological concentrations of CI. The physiological concentration of monomeric CI in a lysogenic cell is 200 nM (6, 17, 30, 31). We confirmed the cooperative role of O_L3 and O_R3 in P_RM repression. Additionally, we also discuss that it is the difference in loop configuration that modulates activation or promotes repression.

Results

We used our in vitro transcription system containing supercoiled DNA templates with the prophage promoters P_R, P_L, and P_RM to investigate the autoregulation of CI by DNA looping. We designed and constructed the templates in order to produce in the same assay the following discrete RNA sizes: P_R, 117 nt; P_L, 167 nt; and P_RM, 212 nt (Fig. 2A). The distance between the midpoint of O_L3 and O_R3 is 392 bp in most DNA templates with an integration host factor (IHF) binding site and an “up” element between O_L3 and the T_imm terminator (Fig. 2A). The in vitro transcription results from the phage promoters in the presence of increasing concentrations of total CI monomers in a supercoiled DNA template (pDL944) with wild-type operators as shown in Fig. 2 B and C. As expected, CI efficiently repressed the lytic promoters, P_L and P_R; half-maximal repression (0.5-fold) of P_L and P_R occurred at 50 nM CI. P_RM was maximally activated (more than ninefold) at 80 nM CI; transcription from P_RM was repressed at higher CI concentrations (Fig. 2 B and C). Half-maximal activation and repression of P_RM occurred at 50 and 320 nM CI, respectively. The observed regulatory pattern was qualitatively similar to previous in vitro and in vivo results, where P_RM was maximally stimulated 10-fold with respect to its basal level (1, 2, 8).

Fig. 2.

Regulation of P_R, P_L, P_L2, and P_RM by CI. (A) Map is drawn to scale with the O_L and O_R regions in black rectangles, promoters (P_R, P_L, P_L2, and P_RM). The hatched boxes show the terminators (T1T2_ter, Timm_ter, and rpoC_ter). An IHF binding site and an “up element” are located to the right of O_L3 (42, 43). Partial rexB and cI genes are shown upstream of P_RM. The distance between the centers of O_L3 and O_R3 is 392 bp. (B) RNAs made from wild-type plasmid in A with increasing total CI monomer concentrations (nanomolar). RNAI transcripts are control RNAs. P_L2, a minor promoter with its transcription start point located 42-bp upstream of that of P_L, was also repressed by CI (42, 43). The physiological significance of P_L2 is not known. (C) Quantification of the transcripts in B. The P_R, P_L, and P_RM transcripts are normalized to RNAI and to the amount of transcripts in the absence of CI (lane 1) for each promoter. Therefore, the amount of each transcript made in lane 1 is taken as 1.0. The relative amount of transcripts (y axis) is plotted against the total CI monomer concentration (x axis). The x axis has a break from 3–4 nM CI and a log scale after the break.

The intrinsic promoter strength of P_RM was very low in the pDL944 template (Fig. 2B, lane 1). To increase the sensitivity of P_RM in our assays, we used an “up mutation” (prmup-1, henceforth called P_RM1), which changes the -35 element of P_RM from to at position -31, creating plasmid pDL985 (1, 16). The increase in the basal activity of P_RM1 relative to P_RM without CI was sevenfold (Fig. 3A, lane 1). At 160 nM CI, P_RM1 was maximally stimulated threefold, which is in agreement with three- and fourfold activation of P_RM1 in vivo (1, 2, 16, 32, 33) (Fig. 3 A and C). Half-maximal activation and repression occurred at 60 and 500 nM CI, respectively. At very high concentrations of CI, P_RM1 was repressed below its basal level (lanes 11–12). Thus, the regulatory effect of CI on the transcripts of P_RM1 was as in the WT P_RM template except that P_RM was activated 10-fold, whereas P_RM1 was activated threefold (Figs. 2 and 3). Both promoters were repressed at high CI concentrations.

Fig. 3.

Effect of DNA spacer length on P_RM1 regulation. (A) RNAs made from P_RM1 plasmid (392-bp spacer). (B) RNAs from P_RM1 with 2.3-kbp spacer. (C) Quantification of RNAs from P_RM1 with 392-bp (A) and 2.3-kbp spacers (B).

The WT distance between O_L3 and O_R3 is 2.3 kbp. To substantiate the results described above and correlate them to WT DNA length, we tested a P_RM1 template, pDL1133, containing the natural length of 2.3 kbp between O_L3 and O_R3 (with rexB, rexA, IHF site, and up element) (Fig. S1 A and B). We inserted an additional transcription terminator (rpoC_ter) downstream of P_RM1 in the cI gene to reduce the size of P_RM1 transcript to 96-nt from its natural length of approximately 2,136 nt, which would have been difficult to detect in our RNA gels (Fig. S1 A). P_RM1 was activated 3.5-fold at 80 nM CI before being repressed at higher CI concentrations (Fig. 3 B and C). Half-maximal activation and repression occurred at 40 and 350 nM CI, respectively. Above 600 nM of CI, P_RM1 was repressed below its basal level of expression. The decrease in O_L ∼ O_R spacing did not significantly change P_RM autoregulation. In the other experiments, we studied P_RM regulation using P_RM1 DNA templates with a 392-bp separation between O_L3 and O_R3. We believe that the use of P_RM1 and shortened O_L ∼ O_R distance in the DNA template would not hinder our goals of studying P_RM regulation. We also assumed that an “antiparallel” loop, and not a “parallel” loop, is the preferred loop geometry (34, 35). Though there is a possibility that with longer operator spacing, there might be some form of parallel loops, previous data show that a short distance favors the antiparallel orientation in a DNA loop (34, 35).

Effect of Each Operator on P_RM1 Regulation.

To investigate the role of each of the six operators in P_RM1 regulation, we inactivated each operator by changing the conserved to sequence in the consensus half-site to remove the operator recognition for CI (Fig. 4A) (36). This 2-bp sequence is important for CI binding as determined by genetic studies (1). Structural studies showed that Lys-4 and Asn-55 of CI contact the G residues at positions -6CG and -7CG of O_R1 or O_L1 (12, 37–40). Both bases are important for CI recognition because changing of either base resulted in a large positive value of ΔG (12, 38, 39). Such mutations in all six operators at the same time prevented both activation of P_RM1 and repression of P_R and P_L by CI, confirming that the mutant operators are nonfunctional (Fig. S2).

Fig. 4.

Effect of operator mutations on P_RM1 regulation. (A) Alignment of operators’ sequences showing the consensus (1–8) and nonconsensus (1′–8′) half-sites. The axis of symmetry passes through the ninth base pair. The conserved penta site (CACCG) is underlined. A change of CC to AT inactivates each operator site. (B) Quantification of P_RM1 RNAs made from templates containing O_R operators (Fig. S3 A–D). (C–E) Quantification of P_RM1 RNAs made from templates with mutations in O_L operators (Fig. S3 E–H). The data for the and constructs are shown in panels C–E, which are separated for clarity.

First, we studied the contribution of individual O_Roperator elements in P_RM1 regulation by CI on templates containing operators.

(i) Mutation in O_R1 is expected to stabilize a CI dimer at weak O_R3 because of a cooperative interaction with CI bound at O_R2 even at low concentrations, which increases the affinity of O_R3 by fivefold (16, 31). Such binding should prevent activation and enhance repression of P_RM. In agreement, we found that inactivation of O_R1 resulted in repression of P_RM1 even at 80 nM of CI and no activation at any CI concentrations (Fig. 4B and Fig. S3B). We propose that this strong repression of P_RM1 was due to both O_R2 ∼ O_R3 cooperativity and the interactions of the CI tetramers at O_R2 ∼ O_R3 and O_L1 ∼ O_L2 by a shift of CI partners changing to a different register because in an O_R1^- template, P_RM1 repression occurred at higher CI concentrations relative to P_RM1 repression in an template (Fig. S4).

(ii) Because O_R3 is only involved in P_RM repression, mutation in O_R3 is not expected to affect the activation of P_RM. Indeed, the P_RM1 activity was stimulated to a maximum level of fivefold at 160 nM CI in the O_R3 mutant template and no repression was observed at higher CI concentrations as expected (Fig. 4B and Fig. S3C).

(iii) We expected that mutation of O_R2 would eliminate P_RM1 activation, but we observed a small amount of P_RM1 activation (1.7-fold) in the O_R2^- template at 240 nM CI before the repression brought about by O_R3 occupation at higher CI concentrations (Fig. 4B and Fig. S3 A and D). Meyer et al. also showed a low level of P_RM activation (ca. twofold) in O_R2^- mutants (virC23 and v1) (1, 10). They suggested that this low level of activation was due to CI binding to O_R1 because, in an O_R1 mutant, the activation was eliminated. We propose that this activation may be due to some level of CI binding to the defective O_R2 site because of cooperative interactions with a dimer bound at O_R1 at high CI concentrations, as suggested previously (41) to explain the rescue of CI binding to a mutated O_R3 site. Incidentally, we also investigated whether the observed P_RM1 repression in the O_R2^- template is because of a DNA loop formed by a cooperative interaction between CI bound to O_R1 and O_R2^- that aids an interaction between CI at O_L3 and O_R3. A DNA loop is involved in P_RM1 repression because the repression was eliminated in the double mutant (Fig. S5).

The above experiments performed in the presence of the O_L region and, therefore, under conditions that permitted DNA looping could assist CI binding at O_R3 to elicit P_RM1 repression at lower CI concentrations than in the absence of O_L. To confirm the role of the O_L region and DNA looping in the regulation of P_RM1, we first tested a template in which all three O_L elements were mutated (triple mutant, ) to eliminate DNA looping. In contrast to the wild-type () template where half-maximal repression of P_RM1 occurred at 500 nM CI, complete repression of P_RM1 in did not occur even at 800 nM CI (Fig. 4C and Fig. S3E). At this CI concentration, P_RM1 was repressed in relative to by threefold and this difference increased to fivefold at 1,600 nM CI. This weak repression of P_RM1 in probably reflects the low intrinsic binding of CI to O_R3 in the absence of DNA looping (Fig. 4C and Fig. S3E). In , P_RM1 was repressed below its basal level at high CI concentrations, whereas in , P_RM1 repression level was above the basal level. In addition, needed less CI than for activation, suggesting that DNA looping enhances activation. This activation became evident when O_R3 was mutated to eliminate the repression of P_RM1 (see below). These results confirm that DNA looping enhances P_RM activation and is essential for P_RM repression at just above physiological CI concentrations.

Next, we assayed the contribution of individual O_L operator elements in P_RM1 regulation by CI on templates containing operators.

(i) Mutation in O_L1 reduced activation and repression of P_RM1 with respect to the WT operator (Figs. 4C and Fig. S3F). Half-maximal activation of P_RM1 occurred at 60 nM CI, and full P_RM1 repression did not occur even at 800 nM CI. To explain this result, we suggest that mutation in O_L1 allows the CI tetramer at O_L2 ∼ O_L3 to interact cooperatively with a CI tetramer at O_R1 ∼ O_R2 by shifting the register, forming an octamer that mediates DNA looping (Fig. S3F). The O_L2 ∼ O_L3/O_R1 ∼ O_R2 loop configuration slightly reduced the activation level of P_RM1 by an unknown mechanism compared to the O_L1 ∼ O_L2/O_R1 ∼ O_R2 loop configuration. In the former configuration, CI at O_L3 is not available to stabilize a CI dimer at O_R3. Therefore, above physiological monomeric CI concentration of 200 nM, P_RM1 should be activated but not repressed fully. Although half-maximal activation of P_RM1 occurred at the same concentration (60 nM) as with the WT template, the maximum stimulatory effect of CI was reduced (Fig. 4C). We infer that the O_L2 ∼ O_L3/O_R1 ∼ O_R2 loop configuration cannot support full activation; the mechanistic basis for this effect remains to be investigated.

(ii) In an O_L2 mutant, P_RM1 repression was reduced to the same level observed in , suggesting that O_L2 is also needed for normal P_RM1 repression (Figs. 4D and Fig. S3G). Half-maximal activation of P_RM1 occurred at 80 nM CI like O_L1^-, and 1,600 nM CI was unable to repress P_RM1 completely. In the absence of CI binding at O_L2, we expected that the octameric loop complex of CI between O_R1 ∼ O_R2 and O_L1 ∼ O_L2 would not form and the cooperative interaction between CI dimers bound to O_L3 and O_R3 would be less efficient or ineffective.

(iii) In the O_L3 mutant like O_L2^- and , repression of P_RM1 at 800 and 1,600 nM CI was three- and fivefold weaker than in WT , respectively. Half-maximal activation of P_RM1 occurred at 80 nM as in . Why there is no activation enhancement of an O_L1 ∼ O_L2 and O_R1 ∼ O_R2 mediated octameric loop on O_L3^- template is unknown. We propose that, in the absence of a stabilizing interaction due to CI binding to O_L3, O_R3 occupancy by CI occurs only at very high CI concentrations due to poor intrinsic affinity and no cooperative effect with O_L3 (Fig. 4E and Fig. S3H). The O_L2 operator is essential for octamer-mediated loop formation, which juxtaposes a CI bound to O_L3 with a CI bound to O_R3 in the antiparallel loop orientation. Our results also suggest that formation of an octameric loop by itself does not lead to P_RM repression; an intact O_L3 must be available for interaction between CI at O_L3 and O_R3 as was interpreted from previous genetic experiments (20). These results confirm that efficient P_RM autorepression at physiological concentrations by CI requires DNA looping with a tetramer of CI between O_L3 and O_R3 plus the octamer of CI between O_L1 ∼ O_L2 and O_R1 ∼ O_R2 (20).

Finally, in order to investigate the contribution of looping to P_RM1 activation, we compared the effect of increasing CI concentrations on transcription from and O_L3^-O_R3^- templates (Fig. 5 and Fig. S3 I and J). In such templates, P_RM1 activation is no longer obscured by the repression mediated by CI bound to O_R3, and in the template, no looping occurs (Fig. S3I). The O_R3 mutation is used to prevent any P_RM1 repression, thereby allowing activation to occur at its maximum level. We found that, at high CI concentrations, P_RM1 was activated up to a maximum of fivefold in the template (nonlooping) and eightfold in the O_L3^-O_R3^- template (looping), (Fig. 5 and Fig. S3 I and J). We argue that this 1.6-fold increase in P_RM1 activation in the O_L3^- compared to template is caused by formation of an octamer among repressor dimers bound to O_R1 ∼ O_R2 and O_L1 ∼ O_L2. It was previously shown in vivo that DNA looping can enhance P_RM activation from two- to fourfold (27, 28). Recent in vivo reporter assays on prm240, a down-promoter mutation of P_RM, template showed that DNA looping can enhance P_RM activation by fivefold (41). It has been proposed that the stimulating effect of DNA looping on the activation of P_RM transcription is attributable to a sterically feasible interaction between the α-carboxyl terminal domain (α-CTD) of RNA polymerase at P_RM and an up element located immediately rightward of O_L3 (42, 43). Due to an antiparallel loop configuration, the up element is brought into proximity to contact the α-CTD.

Fig. 5.

Stimulatory effect of DNA looping on P_RM1 expression. Quantification of P_RM1 RNAs made from looping (O_L3^-O_R3^-) and nonlooping () templates (Fig. S2 I and J).

Role of Cooperative Binding in Autoactivation.

To determine the contribution of CI bound at O_R2 on P_RM1 autoactivation in the absence of other regulatory influences, and we used an template, which should eliminate the effects of DNA looping, autorepression, and CI cooperative binding toO_R1 ∼ O_R2. In the presence of the O_R2 operator only, half-maximal activation of P_RM1 occurred at 600 nM CI, which was 10-fold higher than 60 nM CI required for half-maximal activation of P_RM1 on WT operator template (Fig. 6 and Fig. S3K). No P_RM1 repression was seen, as expected from the absence of CI binding to O_R3. The absence of cooperativity can explain the higher CI concentration required for P_RM1 activation and reflects the contribution of only O_R2 in P_RM1 activation of CI without any cooperative binding.

Fig. 6.

Effect of CI binding to O_R2 on P_RM1 expression. Quantification of P_RM1 RNAs made from control DNA () in the presence of WT CI and mutant CI (G147D) and (D197G). RNAs from DNA in the presence of WT CI (Fig. S2 K–M).

To confirm that enhancement of activation at physiological CI level is due to cooperative CI binding, we tested the effect on P_RM1 transcription of two noncooperative CI protein mutants (G147D and D197G; refs. 44–47) using a WT template. Both CI mutants bind to the operators as dimers and are defective in adjacent and distal cooperative interactions between CI dimers (46, 48–51). The crystal structure of the CI D197G: O_L1 complex revealed important information about cooperative binding to adjacent sites (52). Half-maximal activation of P_RM1 in G147D and D197G were observed at 400 and 350 nM mutant CI, respectively (Fig. 6 and Fig. S3 L and M). P_RM1 repression was not observed in the presence of G147D and D197G proteins, but the observed P_RM1 activation corresponds to a CI dimer bound to O_R2 in the absence of adjacent mutant operators. As expected, in the absence of cooperative binding to adjacent and distal sites, only the intrinsic binding of CI to O_R2 contributed to the low level of activation. In the absence of cooperativity between mutant CI dimers, the O_L3 and O_R3 repression of P_RM1 at lower CI concentrations did not occur. The two approaches (CI mutants in WT operator and WT CI in mutant operator yielded highly consistent results with one another and allowed quantification of the contribution of O_R2 alone in P_RM1 activation.

Discussions

Our study confirmed several predictions of previous investigators discussed above about the mechanism of autoactivation and autorepression of P_RM by measuring in vitro transcription on supercoiled DNA templates containing different combinations of operator mutations. Our transcription data support the model that autoactivation requires CI binding to O_R2, and that activation is stimulated by cooperative dimer–dimer and tetramer–tetramer interactions between CI molecules at O_R2 and those bound to other operator subsites. First, cooperative interactions of CI not only occur between O_R1 and O_R2 in cis, as reported previously, but also between O_R1 ∼ O_R2 and O_L1 ∼ O_L2 in trans, resulting in individual site occupancies at lower CI concentrations as judged by their effects on P_RM activation. The trans-cooperativity by octamer formation requires adjacent CI binding sites between O_R1 ∼ O_R2/O_L1 ∼ O_L2 or O_R1 ∼ O_R2/O_L2 ∼ O_L3 and not O_R1 ∼ O_R2 and O_L1 ∼ O_L3 (Fig. S3 F and J). Similar trans-cooperativity between O_R3 and O_L3 is inferred from P_RM repression at lower CI concentrations. Cooperative interactions lead to DNA looping when they occur between CI at O_L and O_R sites. P_RM is autorepressed by the binding of CI to O_R3 as reported previously (1–3), and this is stimulated by dimer–dimer and tetramer–tetramer interactions that occur through DNA looping. Our results are in agreement with Revet et al. (18) that looping per se strengthens the binding of CI to the suboperators (Fig. 5). Our work provided quantitative evaluation of the contribution of each operator to autoregulation.

Autoactivation.

Given that the presence of O_L promotes DNA looping, we demonstrated in vitro that looping further enhances activation. In vivo, a twofold activation of P_RM by an octameric loop was observed by Anderson and Yang (27, 28), although Little and Michalowski showed a substantial increase (ca. fivefold) in P_RM by looping on prm240, a weak P_RM variant (41). To characterize the conditions that affect activation, the O_R3 site was mutated to alleviate repression of P_RM1. Comparison of the O_L3^-O_R3^- and the templates revealed that looping enhances the activation of P_RM1 by 1.6-fold at maximum activation levels. The mechanism of enhancement of P_RM activation by an octameric DNA loop formation may simply be by the associated tighter binding of CI to O_R2.

Autorepression.

We showed that, in a template with WT operators and promoters, the repression of P_RM commenced above 80 nM CI and was almost fully repressed above 400 nM. The physiological concentration of monomeric CI in a lysogenic cell was estimated to be approximately 200 nM (6, 17, 30, 31). Therefore, in our system, repression of P_RM was observed above physiological conditions. When we tested the regulation of P_RM1 by CI, only partial repression of P_RM1 occurred in the absence of O_L, even at 1,600 nM CI, implying that the binding of CI to O_R3 is not sufficient for complete repression of P_RM1. However, when O_L was present, complete repression was observed, confirming that autorepression requires DNA looping, as was suggested from in vivo experiments (19). It has been proposed that an interaction between CI dimers at O_L3 and one at O_R3 helps CI binding to the latter, but only when looping has occurred with CI bound to O_L1 ∼ O_L2 and O_R1 ∼ O_R2 (19). CI has higher affinity for O_L3 than for O_R3 (11, 13, 15, 53). Our in vitro transcription experiments using various combinations of O_L mutants confirmed the involvement of O_L3 in P_RM repression but also showed the involvement of O_L2 in efficient O_R3 mediated P_RM1 repression by CI. In templates, the presence of an O_L3^- or O_L2^-, but not O_L1^- mutation failed to show P_RM repression at lower CI concentration (Fig. 4 C–E and Fig. S3 A–D). The results further suggest that the cooperative help of CI from O_L3 to CI at O_R3 can only occur if CI also binds to the adjacent element O_L2. Note that (i) the involvement of the O_L element toward P_RM repression is indirect and through O_R, (ii) DNA looping can occur by octamer formation at O_L1 ∼ O_L2 and O_R1 ∼ O_R2 or O_L2 ∼ O_L3 and O_R1 ∼ O_R2.

In summary, our results confirm that DNA loops in prophage λ are likely to contain, in the absence of mutations, four CI dimers (two CI at O_R1 ∼ O_R2 and two at O_L1 ∼ O_L2 forming an octamer) or six CI dimers (an additional dimer pair connecting O_R3 and O_L3). The octamer-mediated loop is less stable on linear DNA (25). We show that looping enhances autoactivation of P_RM1 as long as O_R3 does not become occupied. We confirmed that the loop that enhances the autoactivation is the octamer-mediated, thermodynamically less-stable loop. We also show that, at physiological CI concentrations, looping is essential for P_RM repression but only the stable octamer plus tetramer-mediated form can do so.

Transcription of the cI gene from the P_RM promoter in a lysogen occurs at multiple levels: (i) basal transcription; (ii) activated transcription by CI without any DNA loop; (iii) enhanced activated transcription by a loop comprising four CI dimers; and (iv) repression of basal transcription by a loop containing six CI dimers. This polymorphic behavior of P_RM regulations by a single protein enables the lysogen to maintain a steady but very low level of free CI in a λ lysogen. This low level may allow fast degradation of free CI by SOS, which is induced by RecA, a coprotease (54, 55), and thus easy switching of the prophage to a lytic state. It is interesting to notice that the maintenance of a biologically critical level of CI involves formation and interconversion of two topographically different forms of a loop.

Materials and Methods

Proteins.

E. coli RNA polymerase (1 unit/μL) was from USB/Affymetrix. WT CI and mutant (G147D, D197G) proteins were purchased from Protein RST. They were more than 98% pure as estimated by Coomassie staining of SDS gels. Modified Lowry Protein Assay was used to determine the total CI monomer concentrations.

Plasmid Constructions.

Details of plasmid constructions are given in the SI Text. The plasmids used in this study are listed in Table S1. Table S2 contains the primers used in the construction of WT plasmid pDL944 as described in Fig. S6.

In Vitro Transcription Assays.

In vitro transcription reactions were performed as described (56). Bacteriophage λ repressor CI (20–1,600 nM) was added to supercoiled DNA templates (4 nM) and the different promoters were transcribed by 20 nM RNA polymerase in the presence of nucleoside 5'-triphosphates and 5 μCi [α-³²P]UTP (1Ci = 37 GBq) (3,000 Ci/mmol). The RNAI transcripts present in the plasmids (106 and 108 nts) were used as internal controls to quantify the relative amount of transcripts from P_R, P_L, and P_RM (57). The relative amount of transcripts in the presence of CI was normalized to that in the absence of CI. P_RM transcripts migrated as 212 nt, P_L as 167 nt and P_R as 117 nt on templates where O_R3 is separated from O_L3 by 392 bp. In vitro, we observed two rounds of transcription in our assays (58). Details of in vitro transcription assay are described in the SI Text.