|Home | About | Journals | Submit | Contact Us | Français|
The non-lambdoid coliphage 186 provides an alternative model to the lytic-lysogenic switch of phage λ. Like λ, the key switch regulator, the CI repressor, associates to octamers. Unlike λ, the lytic promoter (pR) and the lysogenic promoter (pL) are face-to-face, 62 bp apart and are flanked by distal CI binding sites (FL and FR) located ≈ 300 bp away. Using reporter and footprinting studies, we show that the outcome, but not the mechanism, of regulation by 186 CI is very similar to λ. 186 CI stimulates pL transcription indirectly by repressing convergent interfering transcription from pR. However, in the absence of the flanking FL and FR sites, CI bound at pR interacts co-operatively with a weak CI binding site at pL and represses both promoters. FL and FR play a critical role; they assist repression of pR and simultaneously alleviate repression of pL, thus allowing high pL activity. We propose that the 186 switch is regulated by a novel mechanism in which a CI octamer bound at pR forms alternative DNA loops to pL or to a flanking site, depending on CI concentration.
Bacteriophage λ is able to reproduce using two alternative modes of development, lytic and lysogenic, and is able to make efficient transitions between these modes. The underlying regulatory network is an unsurpassed model for gene regulation in a genetic switch (Ptashne, 1992), and one often used as a testing ground in the field of gene network analysis and engineering (see Hasty et al., 2001). Bacteriophage 186 is unrelated to λ at the sequence level (Portelli et al., 1998), yet shares the same lifestyle, being a temperate, integrating, SOS-inducible bacteriophage of Escherichia coli (Woods and Egan, 1974). A comparison of the regulatory networks of 186 and λ should highlight those features important for an efficient genetic switch.
The key regulator in the λ switch is the CI repressor protein, which binds co-operatively as a tetramer to adjacent operator sites at OL and at OR to repress the lytic promoters and activate the lysogenic promoter. Recently, another level of λ CI co-operativity has been described. CI tetramers bound at OR and OL can octamerize, forming a 2.3 kb DNA loop that improves repression of the lytic promoters and also enables partial repression of the lysogenic promoter at the CI concentration found in lysogens (Révet et al., 1999; Dodd et al., 2001; Hochschild, 2002). In this paper, we describe the mechanism of action of the analogous CI protein in 186.
The region of the 186 genome that encodes the lysislysogeny switch is shown in Fig. 1. This region is equivalent to the OR region of λ, but one important difference is that the lytic promoter, pR, and the lysogenic promoter, pL, are arranged face-to-face with their start points 62 bp apart, rather than back-to-back (Dodd et al., 1990). Reporter studies have shown that pR is intrinsically ≈ 10-fold stronger than pL and that convergent transcription from pR inhibits pL transcription some 10- to 20-fold (Dodd et al., 1990; Reed et al., 1997; Neufing et al., 2001). Lysogenic CI levels repress pR strongly and also increase transcription from pL, suggesting that CI may activate pL indirectly by relieving pR’s inhibition of pL. There is also some evidence for CI repression of pL at high concentrations (Dodd et al., 1990).
Despite a lack of sequence similarity, there are strong functional similarities between the 186 and λ CI proteins. 186 CI associates to octamers in solution with free energies similar to λ CI (Senear et al., 1993; Shearwin and Egan, 1996). The organization of the two proteins is similar; the 186 CI protein is composed of (i) an N-terminal DNA-binding domain responsible for DNA recognition and containing a likely helix–turn–helix (HTH) motif; (ii) a central insertion-tolerant linker region; and (iii) a protease-resistant C-terminal domain primarily responsible for self-association (Shearwin et al., 2001). However, unlike λ CI, 186 CI does not appear to have an autoproteolytic activity. 186 CI is not RecA sensitive and is instead reversibly inactivated by the SOS-induced phage protein Tum during prophage induction (Shearwin et al., 1998).
Three DNA regions that bind CI have been located near the pR and pL promoters (Fig. 1; Dodd and Egan, 1996). The highest affinity site is at pR itself, where CI binds highly co-operatively, in an all-or-none manner, to a set of three inverted repeat operator sequences centred at the +1, −21 and −42 positions and thus located on the same side of the DNA helix. CI binds with approximately fourfold lower affinity to two flanking sites (FL and FR), each of which contains a single recognizable operator, at the −327 and the +352 positions of pR. In gel shifts, DNA fragments containing the pR, FR or FL regions give a CI-retarded species with the same mobility, suggesting a similar CI stoichiometry at each site. In DNase I footprints, CI makes extended contacts with DNA adjacent to the primary recognition sequences, suggesting that CI binds to each site as a higher order multimer (Dodd and Egan, 1996).
Here, we use a chromosomally integrated lacZ reporter system coupled with a controlled CI expression plasmid to show that the effects of CI on transcription from pR and pL are very similar to the effects of λ CI on PR and PRM. However, the effects of various mutations in vivo and the results of in vitro footprinting experiments show that the mechanisms of CI autoregulation in 186 are quite different from those used by λ. In 186, proper regulation of lytic and lysogenic transcription occurs through the interaction, apparently through DNA looping, of well-separated promoter and operator elements, including a previously unidentified binding site over pL. We propose a novel mechanism for repressor action in which the distal sites and the pL site compete for the fourth dimer of a CI octamer bound to pR, resulting in alternative DNA loops.
Transcription from pR and pL in vivo was assayed with single copy, chromosomal lacZ operon fusions using the λ-based system of Simons et al. (1987), modified as described by Dodd et al. (2001). 186 CI was supplied to the reporter constructs from pZC320-186cI, a derivative of the low-copy mini-F plasmid pZC320 (Shi and Biek, 1995), in which cI is expressed from the wild-type Plac promoter. A second plasmid, pUHA1, was used to supply Lac repressor and allow IPTG control of CI levels. The cellular levels of CI produced with IPTG in this system and also the CI level in a 186+ lysogen were quantified by Western blotting with a CI-specific antibody (data not shown, see Experimental procedures). This allowed us to estimate that 10, 20, 40, 60 and 100 µM IPTG gave average cellular concentrations of CI monomers of 20, 290, 1100, 1800 and 3000 nM respectively. The concentration in the lysogen was 1100 ± 200 nM CI, equivalent to that produced by the expression system at 40 µM IPTG.
Figure 2A and B shows the response of pR and pL transcription to the range of CI concentrations produced by the IPTG-inducible expression system. The 186 DNA fragment used in the reporters was the HincII–SnaBI fragment carrying cIts and aplam mutations (Fig. 1). The fragment contains no active genes but includes the distal CI binding sites FL and FR. CI effectively repressed pR, with half repression occurring at less than 10 µM IPTG or below 20 nM CI (Fig. 2A). At the lysogenic CI concentration, produced by either the expression system or a 186 prophage, pR was repressed from ≈ 400 units to ≈ 1 unit. Basal transcription from pL was very weak, ≈ 7 units, and was activated by CI to a maximum of ≈ 3.5-fold above basal activity (Fig. 2B). With further increases in CI level, pL activity decreased, reaching a level equivalent to basal activity between 60 and 100 µM IPTG. At the lysogenic CI concentration produced by the plasmid expression system, pL activity was ≈ 15 units, reduced ≈ 40% from its maximal value to a level that was 2.2-fold above basal activity. For unknown reasons, slightly less negative autoregulation was seen when the lysogenic CI level was supplied by a prophage (Fig. 2B).
As expected for its key role in lysogeny, the CI protein enforced a huge shift in the relative activities of the lytic and lysogenic promoters: pR activity was reduced ≈ 400-fold, whereas pL was increased from 7 to 15–20 units, a relative shift in expression of ≈ 1000-fold.
Lambda CI stimulates PRM by contacting RNA polymerase (RNAP) (Kuldell and Hochschild, 1994; Li et al., 1994). It was possible that 186 CI bound at pR, FL or FR might make direct contact with RNAP at pL via a DNA loop. Alternatively, as discussed, CI repression of pR may activate pL indirectly by removing pR’s inhibition of pL. To distinguish between these possibilities, we used the pR− mutation created by Reed et al. (1997), which alters the −35 hexamer of pR and reduces pR activity >100-fold. The mutation does not alter the CI recognition sequences at pR (Dodd and Egan, 1996), and we confirmed with gel shift experiments that CI’s affinity for pR was not affected (not shown). As expected, inactivation of pR improved pL transcription in the absence of CI, in this case ≈ sixfold. However, despite all the CI binding sites being intact, CI was unable to activate pL further (Fig. 2B), showing that CI binding does not directly stimulate pL and supporting the relief-of-interference mechanism.
The pR− mutation eliminates CI activation of pL and allows the clear display of CI repression of pL. The pR− data in Fig. 2B show that repression of pL by CI begins at low CI concentrations, is ≈ 50% at the lysogenic concentration and reaches 90% at close to three times the lysogenic concentration.
The operon fusions show that CI regulates the transcription of its own gene by positively and negatively controlling pL activity. We wondered whether CI might also affect its own expression post-transcriptionally. One possibility is that hybridization of the complementary regions of the pL and pR transcripts affects the stability or translation of the cI mRNA (note, however, that the cI start codon is at the pR −50 position). To test this, we constructed transcriptional plus translational reporters of cI expression with the HincII–SnaBI fragment by fusing the first 143 codons of cI with the lacZ reading frame (Experimental procedures). The experiment shown in Fig. 2B was repeated with these constructs, both pR+ and pR−, but in neither case were the results substantially different from those seen with the operon fusions (not shown). Thus, CI autoregulation appears to be purely transcriptional.
The ability of CI to repress pL was puzzling because the identified CI binding sites lie well away from pL, at positions +62, +83, +104 (the three pR sites), +389 (FL) and −290 (FR). We therefore tested the ability of CI to repress pL in the absence of FL and FR by constructing a pL reporter with a short DNA fragment, extending from the −65 to the +143 positions of pL, and carrying the pR− mutation. The results showed that CI does not need FL or FR to repress pL efficiently (Fig. 3A). (In fact, a comparison of Figs 2B and and3A3A suggests that removal of FL or FR improves repression of pL. Experiments to test this properly are presented later.)
The efficiency of repression of pL and pR on the DNA fragment without FL and FR was quite similar (Fig. 3A), suggesting that the same binding sites are responsible for repression of both promoters. As no independent CI binding site at pL was observed in the in vitro screen of Dodd and Egan (1996), we expected that repression of pL was dependent on CI binding to the operators at pR. To confirm this, the ability of CI to repress pL on a −65 to +68 fragment was tested. This fragment contains only one of the CI operators at pR, which by itself is not able to bind CI in vitro (unpublished results). We saw very little or no CI repression of pL on this fragment (Fig. 3A; in fact, the activity was not significantly different from that seen with the pZC320 control). Thus, the binding sites at pR, 60–100 bp downstream of pL, are required for CI repression of pL.
We imagined three possible mechanisms of CI repression of pL. (i) CI at pR might act as a roadblock to elongating RNAP by binding downstream of pL (Deuschle et al., 1986; He and Zalkin, 1992). (ii) CI bound specifically at pR might initiate extended CI binding to adjacent non-operator DNA, causing non-specific occupation of pL, akin to the ‘silencing’ mechanism described for the P1 ParB repressor (Yarmolinsky, 2000). (iii) There may be a specific but very weak binding site for CI (or a host protein) at pL that becomes occupied only by co-operative interactions with CI bound at pR via a short DNA loop. The latter two possibilities are supported by previous in vitro footprinting experiments on short pR–pL fragments, in which protections extending from pR into pL were observed at high CI concentrations (Dodd and Egan, 1996).
To distinguish these models, two approaches were taken. The first was to alter the helical phasing between pL and the CI binding sites at pR. The short loop mechanism should be sensitive to insertions or deletions of half-turns of DNA between co-operating CI binding sites at pR and pL, whereas the roadblock and silencing mechanisms should be relatively insensitive to such changes. In order to make these constructs, we altered 11 bp just to the right of the pR operators (pR−pL construct, Fig. 3 legend). Figure 3B shows that this substitution had no effect on the repression of pL (compare the wild-type pR−pL and pR−pL constructs). However, insertion or removal of 5 bp between pL and the sites at pR abolished CI repression of pL. Repression was substantially restored by the addition or removal of a further half DNA turn (giving changes of +10 or −12 from wild type).
The second approach was to replace pL with a heterologous promoter to test whether specific sequences at pL are important in its repression by CI. The short loop mechanism requires the presence of weak binding sites at pL for CI (or a host protein), whereas the roadblock and silencing mechanisms do not require specific sequences at pL. We therefore placed the λ PRM promoter at the same location as pL relative to the pR sites and found that PRM activity was unaffected by CI (Fig. 3B). We also constructed pR-PRM derivatives with −12, −5, +5 and +10 alterations in spacing and saw no repression of PRM (not shown).
We therefore concluded that a specific binding site must exist at pL and that this site can repress pL only in combination with CI bound at the pR sites and in a helix face-dependent manner.
DNase I footprinting with pure CI was used to see whether CI alone mediated interaction between pR and pL and to try to locate any CI binding sites involved. The DNA templates were derived from the clones used in Fig. 3B and had either the wild-type pR-pL spacing (pR−pL template) or the 5 bp deletion (pR−[−5]pL template). CI binding to pR was evident on both templates (Fig. 4). On the pR−[−5]pL template, CI effects on DNase cleavage extended towards pL only as far as the edge of the rightmost CI operator at pR. However, on the pR−pL template, alterations in DNase cleavage extended right across the pR–pL interval up to the −12 position of pL (Fig. 4). Previously, we interpreted a very similar extended footprint over pL as non-specific binding (Dodd and Egan, 1996). The result with the pR−[−5]pL template clearly shows that this interpretation was incorrect.
The pattern in the pR–pL interval on the pR−pL template contains regions of strongly reduced DNase cleavage (open triangles, Fig. 4) alternating with regions of high or increased DNase sensitivity (filled circles and triangles), with a periodicity close to 10.5 bp. Such a pattern is indicative of DNA curvature; the minor groove on the inside of the loop becomes or remains resistant to cleavage by DNase I, whereas the minor groove on the outside becomes or remains readily cut (Drew and Travers, 1985; Such et al., 1988). The footprint on the wild-type top strand, taken from Dodd and Egan (1996), is also given in Fig. 4 and shows the expected staggering of the pattern across the minor groove. The pattern is very similar to that seen when λ CI binds to operators five or six DNA turns apart and causes looping of the intervening DNA (Hochschild and Ptashne, 1986). However, the data do not exclude the possibility that CI binds continuously, on one face of the DNA helix, between specific binding sites at pR and pL.
Inspection of the DNA sequence at the pL end of the CI footprint revealed a sequence with a partial match to a type A CI recognition sequence. CI is able to bind two distinct recognition sequences (Dodd and Egan, 1996), apparently using the one HTH motif (Shearwin et al., 2001). The single operators at FL and FR and a pair of operators at another lytic promoter, pB, are inverted repeats of the consensus half-site ATTCAC separated by five Ws (As or Ts; Dodd and Egan, 1996). The sequence centred at the pL −2 position: TTTGGC taataGGGAAT (Fig. 4) has four mismatches (underlined) to this 17 bp consensus. The known type A operators have at most two mismatches, explaining why the site at pL does not bind CI independently. This putative CI operator is centred 63 bp, or six DNA turns, from the centre of the rightmost pR operator, placing this site in an ideal position to interact with CI bound at pR. Binding to this site would place two CI monomers on the same side of the DNA as the six CI monomers at pR, each contacting their DNA half-sites in adjacent DNA major grooves, and locating CI perfectly on the inside of the pR–pL DNA loop, in phase with the DNase I-resistant bonds and out of phase with the DNase I-sensitive bonds. Our initial attempts to confirm this putative CI binding site by mutation have failed, and further mutation analysis is required.
We note that occupation of pR and pL by CI occurred at different concentrations. In Fig. 4, the pR sites were almost saturated at 100 nM CI, but protection of the pL site was not seen at this concentration and was only apparent at 200 and 400 nM CI. In further footprint experiments with the same DNA fragment (not shown), we found that occupation of pR was apparent at 40 nM CI, whereas binding to pL was seen only at 120 nM CI or above.
To investigate the role of the distal CI binding sites on regulation of pL and pR, we altered FL and FR to inactivate CI binding (see Experimental procedures). Loss of CI binding to DNA fragments bearing the FL− and FR− mutations was confirmed by gel shift assays (not shown). We then incorporated the mutations into reporters carrying the HincII–SnaBI fragment. Figure 5 shows the effect of inactivation of each or both flanking sites on the response of pR and pL (in the presence or absence of active pR) to CI supplied from the expression system.
CI repression of pR was more efficient in the presence of the distal sites (Fig. 5A). In the FL−FR− double mutant, repression was significantly worse than wild type. At the lysogenic CI concentration (40 µM IPTG), this difference was approximately fourfold. There was some redundancy in the need for the flanking sites; the presence of either single site restored repression to almost the wild-type level. The simplest interpretation of this effect is that CI bound at FL (at pR −327) or at FR (pR +352) can interact co-operatively with CI bound at pR, presumably via a DNA loop. In this way, FL and FR would function analogously to distant auxiliary binding sites for other repressors, such as Lac (Oehler et al., 1990).
CI repression of pL was also affected by the distal sites, but their action was reversed – their presence inhibited pL repression in the pR− mutant (Fig. 5B). Repression of pL in the FL−FR− mutant was significantly stronger than wild type at all CI concentrations. The largest effect was about threefold at the lysogenic CI concentration. Each flanking site contributed to the effect, but the FL site was more effective than FR; in the FL+FR− mutant, pL was repressed similarly to wild type, whereas the repression in the FL−FR+ mutant was intermediate between wild type and the double mutant.
The opposing effects of the distal sites on repression of pR and pL are illustrated in Fig. 5C. In the absence of FL and FR, the efficiencies of CI repression of pR and pL at low CI concentrations are similar, with half repression of both promoters occurring at 10–15 µM IPTG. At higher CI concentrations, pL is repressed down to 10%, whereas pR goes on to be repressed over 100-fold. In the presence of FL and FR, low CI concentrations are now able to discriminate effectively between pR and pL, with pR half repressed at less than 10 µM IPTG (<20 nM CI) and pL half repressed at ≈ 30 µM IPTG (≈ 700 nM CI). The effect is that, when pR is repressed by over 50%, pL is repressed by only ≈ 10%.
The importance of this ability of CI to discriminate between pR and pL is shown by the effect of the flanking sites on pL activity in the presence of pR activity (Fig. 5D). The flanking sites had a very strong effect on CI activation of pL, with activation all but abolished in the FL−FR− double mutant (maximal 1.5-fold activation). As activation of pL is by repression of pR, the flanking sites’ role in allowing CI discrimination between pR and pL seems to be to provide a ‘window’ for increased pL activity. The ability of each flanking site to aid activation of pL correlated with its ability to alleviate pL repression (Fig. 5B): the presence of FL alone gave pL activation close to wild type, whereas the effect of FR alone was weaker.
Thus, the flanking sites have a very large effect on CI regulation of the switch promoters, playing a major role in the ability of CI to shift transcription from lytic to lysogenic. In the presence of FL and FR, CI produces a 1000-fold shift in relative transcription. In the absence of the flanking sites, the lysogenic CI concentration represses pR only ≈ 100-fold and does not activate pL, so that the relative shift from lytic to lysogenic transcription is reduced by a factor of 10.
The conservation of the pattern of CI repressor regulation of the lytic and lysogenic promoters in the evolutionarily disparate phages λ and 186 emphasizes the importance of this regulation in the lifestyles of these phages. These basic regulatory features can be understood using reasoning previously applied to λ (Johnson et al., 1981; Ptashne, 1992).
In the absence of CI, 186 lytic transcription is ≈ 60-fold stronger than lysogenic transcription (Fig. 2). The ratio is somewhat lower in λ, ≈ 15- to 20-fold (Meyer et al., 1980; our unpublished results), presumably to compensate for the poor translation of λ cI mRNA (Shean and Gottesman, 1992). Strong lytic promoters allow for high levels of expression of lytic genes, whereas lysogenic promoters that are weak in the absence of CI are presumably needed to prevent automatic entry into lysogeny and make such entry conditional on other factors, such as CII. 186, like λ, has a cII gene – a lytic gene necessary for the establishment of lysogeny (Neufing et al., 1996; 2001; Shearwin and Egan, 2000).
Efficient repression of the lytic promoters by CI is expected to minimize the expression of lytic genes toxic to the lysogen. In the lysogenic state, 186 pR is repressed some 200- to 400-fold (Figs 2A and and5A),5A), and we find a value close to 500-fold for λ PR (unpublished results). A previously reported value of 65-fold repression of λ PR in the lysogen (Meyer et al., 1980) is likely to be an underestimate because of high background in the reporter.
We found an ≈ 3.5-fold activation of 186 pL by CI (Fig. 2B) and a 4.4-fold activation of λ PRM by λ CI (Dodd et al., 2001). This positive autoregulation is presumably needed to keep CI levels high enough to maintain stable lysogeny.
Both phages have a mechanism that allows lysogenic CI levels to repress the lysogenic promoter significantly. Loss of this regulation increases the activity of the lysogenic promoter by 2.5-fold in a λ lysogen (Dodd et al., 2001) and would be expected to increase it by twofold in a 186 lysogen (Fig. 2B). In λ, this negative autoregulation significantly limits the CI concentration in the lysogen and is necessary for efficient SOS induction of the prophage; we expect that it functions similarly in 186.
Although the regulatory outcomes of CI action in 186 are similar to λ, and regulation in both phages involves distant CI binding sites, the mechanisms used are different.
In 186, the pL promoter is inhibited by transcriptional interference from the strong, convergent pR promoter. Thus, the low activity of pL in the absence of CI results from a protein binding site – for RNAP – located ≈ 80 bp away. Activation of the pL promoter occurs, again at a distance, by CI binding at and repressing pR to relieve this interference. In contrast, λ PRM activation requires only short-range interactions: PRM is intrinsically weak and is directly activated by CI binding adjacent to the −35 region of the promoter (Kuldell and Hochschild, 1994; Li et al., 1994).
Repression of 186 pL appears to occur through CI binding to a site overlapping the promoter. However, this binding is completely dependent on co-operative interactions with CI bound at pR approximately six DNA turns away. This mechanism is similar in principle to λ, in which OR3 is efficiently occupied only through interaction with OL (Dodd et al., 2001).
In 186, the action of CI at pR and pL is critically affected by the FL and FR sites some 300 bp away. These sites assist CI repression of pR, a similar effect to the improved repression of λ PR in the presence of OL (Révet et al., 1999). This improved repression of 186 pR in turn enhances the activity of pL. Thus, the FL and FR sites are needed to allow low CI concentrations to discriminate effectively between pR and pL and thus favour lysogenic transcription.
A relatively simple and novel model that can explain our results and account for CI regulation of pR and pL in the 186 genetic switch is shown in Fig. 6. At present, this model is speculative, as there are some features for which we have little compelling evidence and there are some contradictions with in vitro data. The model is compared with the current model for regulation by λ CI (Révet et al., 1999; Dodd et al., 2001; Hochschild, 2002) and, although the two schemes are different in detail, they share the use of multiple co-operative interactions involving large repressor multimers.
In the absence of CI, strong transcription from pR inhibits the activity of the weaker, convergent pL (Fig. 6A). We propose that, at low CI concentrations, pR becomes occupied and repressed by a CI octamer (Fig. 6B). Although we have not measured the stoichiometry of CI binding to pR, we think an octamer is likely for the following reasons. First, CI binding to pR would seem to require a complex of at least three CI dimers. This is because CI binds highly co-operatively to all three operators at pR, each operator is comprised of two half-sites (Dodd and Egan, 1996), and each CI monomer has a single DNA-binding domain (Shearwin et al., 2001). Secondly, 186 CI associates with octamers in solution with similar free energies to λ CI (Senear et al., 1993; Shearwin and Egan, 1996), and this octamerization of λ CI appears to be functional in DNA binding (Révet et al., 1999; Dodd et al., 2001). As octamers are not likely to form in solution at the low CI concentrations that repress pR, we expect that the octamer assembles on the DNA. We expect that binding over the six turns of the helix at pR is accommodated by DNA bending and by flexibility of the CI linker region. The free CI dimer on the pR–CI8 complex would then be available to contact other CI operators, such as those at FL, FR and pL, by DNA looping. We have shown that CI at pR can interact with pL in vivo and in vitro in a manner consistent with the formation of a short DNA loop. However, the stoichiometry and structure of the pR–CI–pL complex is unknown, and the proposed operator at pL has not been verified. DNA looping to FL and FR is inferred from the ability of these sites to improve CI repression of pR. Looping between the primary lac operator, O1, and either of the weak O2 and O3 auxiliary Lac operators, located 401 and 92 bp away, improves lac repression by 20- to 40-fold (Oehler et al., 1990). The ability of 186 CI to mediate DNA looping is supported by our recent finding that repression of λ PR by a chimeric protein with the N-terminal DNA-binding domain of λ CI and the C-terminal oligomerization domain of 186 CI is improved by the presence of λ OL 3.8 kb away (unpublished results).
We propose that FL and FR, being stronger CI binding sites than the site at pL, will effectively compete with pL for the free dimer on the pR–CI8 complex (Fig. 6B). With the pR–CI8 complex looped to FL or FR, the operator at pL would be unoccupied, leaving pL unrepressed and free from interference from pR, giving maximal pL activity (Fig. 6B). Note that, even in the presence of FL and FR, low concentrations of CI repress pL slightly (Fig. 5C). Therefore, we expect that there is always some fraction of the pR–CI8 complex that is looped to pL, preventing pL ever reaching the activity seen in the pR− mutant in the absence of CI (Fig. 2B). The pR–CI complex does not appear to present a significant roadblock to RNAP from pL.
We propose that, at high CI concentrations, FL and FR become occupied by CI multimers independently of pR, breaking the DNA loop with pR and allowing the pR–CI8 complex to loop to pL (Fig. 6C). FL and FR appear to contain only single type A operators, but we believe for several reasons that at high concentrations CI binds to FL or FR as a higher order multimer, probably an octamer, by contacting nearby weak binding sequences (Fig. 6C). First, from the mobilities of complexes in gel shift experiments (Dodd and Egan, 1996), CI appears to bind to FL and FR with a similar stoichiometry with which it binds to pR. Secondly, CI footprints at FL or FR in vitro reveal extensive contacts with adjacent DNA (Dodd and Egan, 1996). Thirdly, a single type A operator placed in a different DNA sequence context binds CI some 20-fold weaker than FL or FR, whereas pairs of type A operators separated by two or three DNA turns bind CI strongly (Dodd and Egan, 1996; Shearwin et al., 2001; unpublished results). Therefore, we believe that there are cryptic, weak CI binding sequences surrounding the FL and FR operators that allow the assembly of a CI octamer at these sites in the absence of the DNA loop to pR. Thus, as the CI concentration increases, the protective effect of the flanking sites is progressively removed, and pL becomes increasingly repressed.
The alternative looping model is qualitatively consistent with the in vivo reporter data on CI regulation of pR and pL and much of the in vitro data. However, there are two in vitro results that do not fit the model. First, in the in vitro DNase I footprinting experiments on the pR–pL DNA fragment in the absence of FL and FR (Fig. 4), occupation of pL by CI occurred at concentrations significantly higher than those required to occupy pR. This result implies that the stoichiometry of the in vitro CI–DNA complex is increasing with increasing CI concentration. The model predicts instead that pR and pL occupation should reach their maximal values at the same CI concentration, consistent with the in vivo results, which show that, in the absence of FL and FR, repression of pR and pL occurs co-ordinately (Figs 3A and and5C).5C). Secondly, attempts to replicate in vitro the action of FL in relieving repression of pL, by showing a direct effect of FL on CI occupation of pL by DNase I footprinting on linear templates, were unsuccessful (not shown). We therefore suspect that DNA looping is not occurring efficiently in vitro. Formation of the DNA loops may require DNA supercoiling, host proteins or some other conditions that our in vitro assays lack. Co-operative interactions between Lac repressor at O1 and O3 (93 bp apart) require supercoiling in vitro and in vivo (Borowiec et al., 1987; Sasse-Dwight and Gralla, 1988). A similar interaction between GalR at OI and OE (113 bp apart) requires supercoiling and the host protein HU (Aki et al., 1996). We note that there is a weak binding site for the DNA-bending host protein IHF between FL and pR (unpublished results). Further testing of the model by in vivo methods is planned.
NK7049 (ΔlacIZYA)X74 galOP308 StrR Su− from R. Simons (Simons et al., 1987) was the host for all LacZ assays. DH5α and MC1061 were hosts for recombinant DNA work. Cells were grown at 37°C in LB (Miller, 1972) with the addition of ampicillin (100 µg ml−1 except 30 µg ml−1 for pZC320) and kanamycin (50 µg ml−1 for pUHA1). λRS45ΔYA (Fig. 7; Dodd et al., 2001) was used for integrating lac reporter constructs into the Escherichia coli chromosome.
Fusions of 186 DNA fragments to lacZ were first created in colE1-based, ampicillin-resistant plasmids (Fig. 7). Four lacZ reporter plasmids were used: pMRR9, pMRR9R, pBC2 (used for operon fusions) and pRS414ΔYA (used for protein fusions). All the plasmids contain four copies of the rrnB T1 terminator upstream of the cloning site to reduce background transcription (Simons et al., 1987) and, except for pRS414ΔYA, have stop codons in each frame between the cloning site and lacZ. The pMRR9 and pMRR9R plasmids also contain the weak trpt transcription terminator between the cloning sites and lacZYA (Fig. 7).
The HincII.22547–SnaBI.23550 cIts.aplam fragment (HS) was ligated into the SmaI site of pMRR9 in either orientation to give pMRR9-HSFL+pR+pL+FR+ (pIBD011) and pMRR9-HSFL+pR+pL+FR+ (pRAS021; the promoter oriented to express lacZ is underlined). The cIts mutation was cItsp (CI:22G→R; Lamont et al., 1993) and the aplam mutation was aplam32 (Apl:32Y→am), created by oligonucleotide-directed mutagenesis as described by Dodd et al. (1993). The HS fragment from pRAS021 was excised using EcoRI and PstI sites in the pMRR9 MCS (multiple cloning site) and inserted into the EcoRI and PstI sites of pBluescript SK– (Stratagene) to give pBS-HSFL+pR+pL+FR+ (pIBD031). The pR− mutation described by Reed et al. (1997), an alteration of the −35 hexamer (TTTACT→CTCGAG) that reduces activity over 100-fold, was introduced into pIBD031 by in vitro resections to give pRAS035. Mutations to inactivate FL or FR were introduced into pRAS035 by the Quickchange oligonucleotide-directed mutagenesis method (Stratagene): FL was altered from 22726-CTTCACTTAATGTGAAT to CCAGCGAGACGTGAGCC, FR from 23404-ATTCACTTTATGTGAAT to AGCGGTGCTAGCAATTA. In vitro resection was used to create the various combinations of the three mutations in the pBS-HS background. The resultant HS fragments were then put back into pMRR9 using the EcoRI and PstI sites.
The HS fragment, containing either or both FL− and FR− mutations, was excised from the appropriate pBS-HS plasmid (above) using the flanking HindIII and BamHI sites and inserted into the same sites of pBC2, orienting the fragment such that lacZ is expressed from pR.
DNA fragments from the 186cIts MaeII.22980 to MaeIII.23183 sites (MM), containing the pL −65 to +143 region, were prepared by polymerase chain reaction (PCR) from pR+pL+ or pR−pL+ templates using primers bearing flanking XbaI sites (#181: GGGTACCTCTAGACGTTGCTCCATCCTAAAGA; and #182:GGGTACCTCTAGAGTAACGATAGGTGCAGGCAC). The DNA was cut with XbaI and inserted into the XbaI site of pMRR9R to give pMRR9R-MMpR+pL+ or pMRR9R-MMpR−pL+. The pL −65 to +68 region was similarly cloned into pMRR9R using primer #182 and another XbaI-flanked primer (GGGTACCTCTAGATTGGCTAAACCCACGCAATT) to give pMRR9R-MMΔpRpL+.
Plasmid pBC2-MMpR−pL+ was created by inserting the XbaI-digested #181/#182 PCR product (above) into the AvrII site of pBC2. Plasmids pBC2-MMpR−pL+ and pBC2-MMpR− [+5pL+ are the same as pBC2-MMpR−pL+ except that the normal pR+15 to +25 (pL+38 to +48) sequence, GATGGCAAGTG, is replaced by CCTAGGCTAGC and CCTAGGAGGAGCTAGC (containing NheI, BseRI and AvrII sites respectively). These plasmids were constructed by cloning PCR fragments generated with primers bearing these sequences and primers #181 and #182, followed by cutting with NheI, AvrII and XbaI, ligation and insertion into pBC2. By resections using these sites, the sequences were then altered to CCTAGC ([−5] construct) and CCTAGGCTAGGAGGAGCTAGC ([+10] construct). The [−12] construct was made by cutting pBC2-MMpR−[+5]pL+ at the AvrII and BseRI sites and using T4 DNA polymerase to fill in the AvrII end and chew back the BseRI end before religating. There was an unexpected loss of the C from the AvrII end, such that, in pBC2-MMpR−[−12]pL+, the normal pR+15 to +30 (pL+32 to +48) sequence, GATGGCAAGTGTTGGC, is replaced by CCTA. Analogous procedures were used to create the equivalent pBC2-MMpR−[*]PRM+ constructs, which carry wild-type λ PRM (−65 to +38) instead of 186 pL.
To construct the cI::lacZ protein fusions, the HS fragment was excised from pIBD031 (FL+pR+pL+FR+) or pRAS035 (FL+pR−pL+FR+), using EcoRI and BamHI sites in the MCS, and inserted into the EcoRI and BamHI sites of pRS414ΔYA to give pIBD041 and pIBD045. This fused codon 143 of cIts to codon 9 of lacZ with three extra codons between the two frames (CIts:M1–V143::GDP::LacZ:9V-end).
The sequences of all mutagenized or PCR-amplified regions of DNA incorporated into plasmids were confirmed, and all cloning junctions were checked by sequencing or by recutting Further details of the cloning procedures are available on request.
Plasmid-based lacZ fusions were transferred to the lacZ reporter phage λRS45ΔYA for insertion into the E. coli chromosome. Plasmid-containing strains (MC1061) were infected with λRS45ΔYA, and blue plaque-forming phage among the progeny were identified and purified on NK7049 on plates containing Xgal (Simons et al., 1987). In the cases in which transfers were from pMRR9- or pMRR9R-based plasmids (lacYA+), these phage were then screened by PCR to check that the recombinants had retained the ΔYA deletion. The resulting λRS45ΔYAlacZ reporter phage (see Fig. 7) were used to lysogenize NK7049, monolysogens being identified by PCR (Powell et al., 1994). We note that the HS reporters would be expected to produce a truncated CIts protein (143 of 192 amino acids); we expect that this protein does not contribute to CI activity, as CI truncations inactivate the protein (Shearwin et al., 2001), the cIts mutation is non-permissive at 37°C, and the protein is expressed at low levels from the single-copy reporter (it was not visible in Western blots).
The wild-type 186 cI gene (186: 22403–23030; beginning 16 bp upstream of the start codon) was amplified by PCR (primers GGAATTCTGaataggttttatcg and CACGGatccaaccgccagcc; 186 sequence is lower case). The product was digested at the flanking EcoRI and BamHI sites and inserted into the corresponding sites in pBlueScript KS+ (Stratagene) to give pRAS1. To make the CI expression plasmid pZC320-186cI (pRAS11), the cI gene was excised from pRAS1 with HindIII and BamHI and inserted at the same sites downstream of Plac+ in the mini-F based, ampicillin-resistant plasmid pZC320 (Shi and Biek, 1995), such that CI is expressed from its natural Shine–Dalgarno sequence. The sequence of the cloned cI gene was confirmed.
To control CI production from Plac+ on pZC320-186cI, Lac repressor was supplied by pUHA1, a p15A plasmid encoding kanamycin resistance and carrying the wild-type lacI gene and promoter, obtained from H. Bujard (Heidelberg University, Germany). The levels of CI produced by the expression system with different IPTG concentrations were quantified by chemiluminescent Western blotting of whole-cell extracts of NK7049(λRS45ΔYA-pMRR9-HSFL+pL+pR+FR+)pUHA1pZC320-186cI, as described by Dodd et al. (2001) for λ CI, but using a rabbit polyclonal antibody raised against purified 186 CI and using purified CI (≈ 98% homogeneous; Shearwin and Egan, 1996) added to CI-less cell extracts as standards. This allowed us to estimate that 10, 20, 40, 60 and 100 µM IPTG gave 16, 226, 869, 1410 and 2390 CI monomers per cell respectively (averages of at least two determinations of two independent extracts; there was no detectable CI without IPTG). In cells in which pZC320-186cI was replaced by a 186 prophage, we found 861 CI monomers per cell (95% confidence limits 712–1010, n = 4). From the generation time of these cultures (35 min), the average cell volume can be estimated as 1.31 × 10−18 m3 (Donachie and Robinson, 1987), so that 1 molecule per cell is equivalent to 1.27 nM.
Kinetic LacZ assays were carried out in 96-well microtitre plates using a plate reader, by an extensively modified method of Miller (1972) as fully described by Dodd et al. (2001). Fresh colonies on selective LB plates were resuspended in LB and used to inoculate 200 µl of LB + antibiotic + IPTG. Dishes were sealed and incubated overnight at 37°C without shaking. For assaying repression of pR, we found that subculturing and a second overnight incubation was needed to dilute away residual LacZ. Overnight cultures were subcultured for assay by diluting 2 µl into 98 µl of fresh medium and incubating at 37°C with rotation to an OD600 of 0.2–1.2 (log phase). Cells were chilled, and then 50 µl of culture was added to 150 µl of TZ8 (100 mM Tris-HCl, pH 8.0, 1 mM MgSO4, 10 mM KCl) +2.7 µl ml−1 2-mercaptoethanol and 50 µg ml−1 polymyxin B. For strong promoters, the culture was first diluted fivefold in LB. Assays at 28°C were started by the addition of 40 µl of ONPG (4 mg ml−1 in TZ8), and the A414 was followed for 1 h. LacZ units were calculated as 200000 × (A414 min−1)/(OD600 × V), with V being the volume (ml) of culture added to the assay. Appropriate background units (obtained from λlac reporter strains with no cloned insert) were subtracted.
DNase I footprinting used the method of Sandaltzopoulos and Becker (1994), with streptavidin-conjugated magnetic beads (Dynal). Footprinting templates were prepared by PCR with one primer labelled with 32P (using T4 polynucleotide kinase) and the other biotinylated. The PCR products were bound to the beads and purified as described by Shearwin and Egan (2000), except that the bead–DNA complexes were resuspended in CIFP buffer [10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.2 mM dithiothreitol (DTT), 5% glycerol, 50 µg ml−1 BSA, 10 mM KCl, 10 mM MgCl2, 2 mM CaCl2]. The biotinylated primer was biotin-5′CCAGCGCCCGTTGCACCACAG (hybridizes in lacZ), and the 32P-labelled primer was TGCCAGGAATTGGGGATC (hybridizes upstream of MCS in reporter plasmids).
DNA diluted in CIFP (45 µl) and 5 µl of purified CI (≈ 98% homogeneous; Shearwin and Egan, 1996) diluted in TEG150 (50 mM Tris-HCl, 0.1 mM EDTA, 10% glycerol, 150 mM NaCl, pH 7.5) were combined at 20°C for 30 min. DNase I (Boehringer; 5 µl of 20 pg µl−1 in 9:1 CIFP–TEG150) was added, followed after 2 min by 55 µl of 4 M NaCl, 100 mM EDTA. Bead–DNA complexes were washed, concentrated, resuspended in formamide loading buffer and analysed by PAGE as described by Shearwin and Egan (2000). Marker sequencing tracks were prepared by dideoxy chain termination sequencing reactions using the labelled primer and the same template DNA used to prepare the footprinting template.
We thank Donald Biek, Hermann Bujard, Thomas Linn and Bob Simons for gifts of plasmids, phage and bacterial strains. We thank the members of the Egan laboratory, Rachel Schubert, Benji Callen, Keith Shearwin and Alison Perkins, for pure CI, CI antibody, plasmids and many helpful discussions. Research in the Egan laboratory is funded by the Australian Research Council.