A functional assay in Escherichia coli to detect non-assisted interaction between galactose repressor dimers

Abstract

Among the Escherichia coli operons repressed from multiple sites on DNA, the galactose operon is unique: its repression requires an auxiliary protein, HU, to assist cooperative repressor binding to two distant DNA sites. Here we show that GalR can still mediate repression from distant sites in an artificial and simplified regulatory region which totally disturbs the organisation of the natural interactions. This simple and unexpected cooperation of a protein incapable of self-association in solution might be involved in regulation of the gal operon. Furthermore, the assay may be generalised to detection of rather weak cooperative interactions between DNA-bound proteins.

INTRODUCTION

The Escherichia coli GalR repressor binds to two operator sites, Oi and Oe, in the galactose operon (Fig. 1A). Both of these DNA sequences are required for full repression of the operon (1). It was originally expected that the observed cooperation would be the result of a simple bridging of the two operator sites by the repressor protein, with looping out of the intervening DNA (1), as observed for some other E.coli operons. However, according to the most recent studies, an auxiliary protein (histone-like protein HU) must assist the interaction between gal dimers in vitro, along with DNA supercoiling (2–4). Two other proteins (the CRP activator and RNA polymerase) also bind to the sequence between the two operator sites and may possibly affect interaction between DNA-bound GalR molecules (5 and references therein).

Figure 1.

The DNA sequence of the regulatory region of the E.coli gal operon (A) and that of the pLacZ plasmid used for construction of the artificial gal operator promoter derivatives (B). The Oi and Oe sites of the gal operon are in pink lettering, as are the XbaI and SpeI restriction sites of pLacZ for introduction of the Oi and Oe operators, respectively. The cAMP–CRP site is in blue lettering. HU binds to the underlined sequence. The ‘Pribnow boxes’ of the promoters are framed in yellow. The coding sequences for β-galactosidase and epimerase are framed in green.

However, some observations suggested to us that non-assisted GalR–GalR interaction might be possible: (i) derepression of gal operon expression is not total in cells depleted of the two genes encoding HU and, moreover, derepression differs in such strains, from 2- to 8-fold (2,6); (ii) gyrase inhibitors cause total derepression of galP2 expression (6); (iii) strains depleted of both HU subunits are rather sickly and spontaneously accumulate secondary mutations (e.g. in the gyrase genes; 7); (iv) HU might stimulate gyrase activity as well as decrease topoisomerase activity (7). Thus, in vivo DNA supercoiling, rather than HU, is an essential component of DNA looping-mediated repression of the gal operon.

Since the use of hup mutants does not lead to clear conclusions, we have used a complementary approach. To determine to what extent HU, as well as the other auxiliary proteins, affect the interaction between DNA-bound gal dimers we have analysed GalR-mediated repression at artificial promoter derivatives.

MATERIALS AND METHODS

Escherichia coli strain CSH50Δ1 [ara, (lac pro)Δ, (galR lysA)Δ, strA] cannot produce β-galactosidase nor the gal repressor. When necessary, this strain was co-transformed with a pLacZ plasmid derivative expressing β-galactosidase under control of the desired combination of gal operators and the pGalR plasmid producing the repressor. Repression of β-galactosidase synthesis can then be followed for this construct. Repression of the lacz gene was followed with this method when the gene was under control of the lac operators and lac repressor (8,9).

The operator sites included in the appropriate plasmids result from annealing of sense and antisense synthetic oligonucleotides (Génosphère, Evry). The minimal 16 bp sequence which was used by Haber and Adhya (1) is flanked by two additional base pairs on each side: ‘Oe’ sequence, 5′-AA.GTGTAAAC.GATTCCAC.TT-3′; ‘Oi ’ sequence, 5′-AA.GTGGTAGC.GGTTACAT.TT-3′.

The pLacZ plasmid is a pACYC184 derivative. It is a low copy number plasmid with less than 15 copies (9,10). It is tetracycline resistant. Its promoter region is detailed in Figure 1B. When required, the 20 bp Oi operator sequence was inserted into the XbaI restriction site and the 20 bp Oe sequence into the SpeI restriction site. The number of operator sequences inserted into either the XbaI site or the SpeI site was determined by PCR with two ‘external’ primers, generating a fragment larger than the region between the XbaI and SpeI sites. The orientation of the inserted operator was determined with another set of primers, the sense or antisense operator oligonucleotide and one of the external primers. Presence of the expected PCR fragment for either the sense or antisense operator primer led to the orientation assignment.

The insertion of 8 bp between the operators of the ←0e←0i construct was performed by filling in of the BglII restriction site (see Fig. 1B) by the Klenow fragment of E.coli DNA polymerase I (Life Technologies).

Insertion of 14 bp between the operators of the same ←Oe←Oi construct was performed by introducing the self-complementary oligonucleotide 5′-GATCTGTCTAGACA-3′ into the BglII site. The reconstituted DNA duplex includes an XbaI site. The XbaI site allows characterisation of the insertion into the BglII site.

The gal repressor coding sequence was copied and amplified from the genome of E.coli strain DH5α by PCR. The pGalR plasmid was built by insertion of the galR gene with its natural promoter region between the EcoRI and AccI restriction sites of the pBR322 cloning vector. This operation preserves the ampicillin resistance gene and removes the tetracycline gene in pBR322. The resulting plasmid is compatible with the pLacZ derivative. Its copy number is greater than 25 (10). It produces sufficient gal repressor to occupy part or all of the gal operator sites carried by the pLacZ derivative since repression of β-galactosidase synthesis is detectable.

The pGalRΔ plasmid is derived from the pGalR plasmid. It carries a non-functional GalR repressor resulting from deletion of codons 15–291 of galR.

β-Galactosidase activities were determined after cell lysis with chloroform and SDS according to the procedure described by Sambrook et al. (10), except that the cultures were grown to an absorption of 1.

RESULTS

The artificial constructs that we have used to analyse gal repressor-mediated repression were derived from a previous study related to lac repressor-mediated repression and oriented heterodimers (9). In these constructs the expression of β-galactosidase was controlled by two lac operators. One of the lac operators was inserted adjacent to the –10 sequence, the other one 102 bp upstream (Fig. 1B). β-Galactosidase expression cannot be activated by CRP protein since there is no binding site for this protein in the promoter region. Furthermore, in these constructs RNA polymerase cannot bind simultaneously to the repressor in a pre-formed repressor–DNA loop since the operators overlap the promoter region.

For this study we replaced the lac operators by gal operator sites. Thus, the lacz gene is under the control of wild-type gal operators. In the present study, for each combination of operators β-galactosidase activity was determined in cells carrying both plasmids, the pLacZ (pACYC184) derivative and the pGalR plasmid (pBR322 derivative), in the absence and presence of the inducer (compare 8,9) d-galactose (40 mM). The repression data are shown in the first and second columns of Table 1.

Table 1. β-Galactosidase activities of gal operator constructs.

	GalR plasmid	GalR plasmid + inducer	GalRΔ plasmid
pLacZ	325.0± 50.4	363.7 ± 10.3	387.6 ± 52.4
Oi→	87.4 ± 15.3	212.1 ± 15.8	316.9 ± 14.9
←Oi	260.6 ± 10.2	597.3 ± 9.1	551.8 ± 56.6
←Oe	539.9 ± 44.3	497.0 ± 20.9	402.3 ± 35.2
Oe→	442.4 ± 62.4	490.0 ± 9.8	478.7 ± 21.8
Oe→Oi→	25.2 ± 2.9	47.1 ± 3.0	254.1 ± 11.0
←OeOi→	24.2 ± 0.9	55.1 ± 4.2	332.3 ± 45.9
Oe→←Oi	32.2 ± 2.7	66.6 ± 2.0	465.7 ± 69.9
←Oe←Oi	39.1 ± 1.0	109.3 ± 10.2	570.3 ± 31.0

The lacz gene is under the control of the indicated combination of operators. The arrows indicate the orientation of the wild-type operator sequence with regard to the start of transcription.

Strain CSH50Δ1 has been transformed with either the gal operator construct and the pGalR repressor plasmid (first and second columns) or the gal operator construct and the plasmid with the repressor deletion (third column). The first column refers to specific activity of β-galactosidase in the absence of inducer for each construct. The second column refers to specific activity of β-galactosidase in the presence of d-galactose as inducer (40 mM). The third column reports specific activity of the specified gal operator construct in the presence of the repressor deletion plasmid, pGalRΔ. Each result is the average of between three and six independent measurements. The indicated error is the mean deviation of the measurements from their arithmetic mean value.

Repression was not totally relieved by inducers, neither d-galactose up to 100 mM nor d-fucose up to 20 mM (data not shown), for the promoter derivatives carrying two DNA sites for GalR (Table 1, lines 6–9, second column, and Table 2, lines 1–3, second column). Similarly, high levels of lac repressor did not allow full induction of the lac operon or of its artificial derivatives (11,12). d-Galactose forms a ternary complex with the gal repressor and gal DNA and the repressor remains on the promoter (13). High levels of gal repressor do not prevent clearance of the promoter from one site (13; our data). Under the same conditions, the repressor is not cleared from two sites (our data), probably because the higher order complex of GalR–GalR–DNA and inducer is more stable than the ternary complex. Therefore, as in previous studies with the lac repressor (compare 8,9), we employed a vector which encodes a defective galR gene, pGalRΔ (Tables 1 and 2, third column), instead of plasmid pGalR and the d-galactose inducer.

Table 2. β-Galactosidase activities of the gal operator ←Oe102←Oi construct with insertions of 8 and 14 bp between operators.

	GalR plasmid	GalR plasmid + inducer	GalRΔ plasmid
←Oe102←Oi	39.1 ± 1.0	109.3 ± 10.2	570.3 ± 31.0
←Oe110←Oi	287.1 ± 8.7	636.2 ± 27.0	492.2 ± 90.6
←Oe116←Oi	33.7 ± 1.7	53.3 ± 5.7	530.0 ± 35.1

Designations are as in the legend to Table 1.

As expected, activity was only weakly repressed by GalR on the two promoter constructs carrying the Oi operator (Table 1, lines 2 and 3). Moreover, repression was not significantly affected by GalR on the two promoter derivatives carrying the Oe operator (Table 1, lines 4 and 5). However, if an additional operator was inserted into the single operator derivatives, repression was strongly increased. Depending on the orientation of the operators, repression was 10- to 15-fold (Table 1, lines 6–9).

In fact, Lanzer and Bujard (14) have shown that the efficiency of repression is mainly determined by the position of an operator in the promoter region and the kinetic parameters of the RNA polymerase–promoter interaction. The Oi site overlaps the start site of transcription. Therefore, it is not surprising that the orientation of Oi can have an effect on repression as well as on promoter strength.

At this stage, we have shown that cooperation between the two sites in vivo does not require HU nor the CRP activator or the RNA polymerase. If the gal repressor is tetramerised in vivo, leading to loop formation and repression of transcription, insertion of a half-integral number of helical turns between the two operators is expected to impede tetramerisation and relieve repression. In contrast, insertion of a spacer with an integral number of helical turns will restore repression. Several studies have shown that the helical repeat of DNA in vivo differs from the helical repeat of unconstrained DNA in vitro (10.5 bp/helix turn) and varies between 11.1 and 11.7 bp (15–17). DNA supercoiling is in part responsible for this (18). We reasoned that the two operator sites were favourably positioned for DNA looping in the gal operon, i.e. aligned on the same face of the DNA. If this is the case, the two sites are separated by an integral number of helical turns. Since the two sites are separated by 114 bp, the resulting helical pitch would be 11.4 bp.

If the sites are aligned (‘in phase’) for 114 bp, they should be out of phase when adding or removing half of the helical repeat, i.e. 5.7 bp, to 114 bp (giving 108.3 and 119.7 bp) and again in phase when adding or removing the helical repeat, 11.4 bp, to 114 bp (giving 102.6 and 125.4 bp). The distance between the two operators in the ←Oe←Oi construct is 102 bp (close to the calculated 102.6 bp distance). This construct is renamed ←Oe102←Oi in Table 2. For technical reasons (see Materials and Methods) we inserted 8 bp between the operators of this construct. The distance between the operators becomes 110 bp (or 9.6 turns). The two sites should be out of phase in the ←Oe110←Oi construct. Insertion of the minimal 14 bp self-complementary oligonucleotide leads to 116 bp (10.1–10.2 turns for a helical repeat of 11.4 bp). The sites should again be in phase in the ←Oe116←Oi construct.

The data are shown in Table 2. When the sites are separated by 102 bp the level of cooperativity is high. Repression is effectively relieved for 110 bp and restored for 116 bp. Thus, there is a clear effect of orientation of the two sites on the DNA. The local helical repeat which fits with the observed data (a high level of cooperativity for the natural distance of 114 bp and the artificial 102 bp and 116 bp distances; loss of cooperativity for 110 bp) is ~11.5 bp, in the range of values found by other groups (15–17).

DISCUSSION

The wild-type galactose operon is repressed at two operator sites, Oi and Oe. The upstream operator, Oe, was discovered first. The question of how this operator could repress transcription upstream of the operon was resolved when the second operator, Oi, was found, located within the first structural gene of the operon (19). To explain the involvement of the two sites in repression, Adhya and co-workers postulated several mechanisms. Interaction between the two gal dimers bound at the operator sites was one of them (19). The two operators are required for full repression of the operon in vivo and function synergistically, as shown by successive destruction of the sites (1). In fact, two lac operators were substituted for the gal operators, showing that gal DNA per se does not impede DNA looping, as observed by Brenowitz and co-workers (20).

However, interaction of gal dimers bound at the operator sites could never be proved in vitro in the absence of auxiliary proteins. Our data indicate that the gal repressor can act cooperatively outside the context of the wild-type galactose promoter in vivo. The promoters have totally arbitrary sequences and exclude site-specific protein binding. In particular, insertion of spacers between the operator sites, with subsequent modification of the helical phasing of the operators and an effect on repression, supports the existence of protein–protein interactions.

More precisely, when HU binds to the gal promoter, a complex with a precise architecture is formed (3). HU is specifically located in the –10 to +30 promoter segment (3). It is unlikely that such a complex is formed with our constructs. The precise geometry of the GalR–HU–gal DNA complex would have to adapt to the various arbitrary sequences and distances found in these constructs.

Apart from HU, other proteins have been said to affect the interaction between gal repressors. Transcription of the gal operon is initiated from two overlapping promoters, P1, in the presence of the activating complex cAMP/CRP, and P2, in its absence (Fig. 1A).

As far as the gal operon is concerned, several studies (5 and references therein) have shown that the gal repressor does not inhibit CRP/cAMP and RNA polymerase binding to the gal regulatory regions. Thus, regulation of open complex formation at the gal promoters would not result from competition between the gal repressor, CRP/cAMP and RNA polymerase for binding at the gal operon regulatory region, but would instead result from interactions of the three proteins during formation of a nucleoprotein complex on the gal DNA fragment. This nucleoprotein complex would be favoured by DNA bending induced by the constituent proteins.

Our data indicate that CRP or RNA binding to the loop are not essential for GalR dimer cooperation in vivo. With the same constructs we have previously shown that the observed cooperation between two lac operators closely positioned in the promoter region, like the wild-type lacO1 and lacO3 operators, is effectively due to their interaction and not to some indirect effect of CRP protein (8), exclusion of the repressor from its binding site (21) or assistance by CRP protein (22).

Various models have been proposed to explain how the auxiliary proteins facilitate GalR repression on gal DNA (2,5). From the most recent works by Adhya and co-workers it appears that the structure which is formed is closer than initially thought to that found in related systems: HU does not contact GalR (3) and the two dimers are capable of specific interactions (as we have found), since mutations in the GalR gene abolish cooperativity (23). Since HU does not bind to gal DNA in the absence of GalR (3), HU binding may be dependent on direct GalR–GalR cooperation and may stabilise an interaction which is weaker in its absence.

In this case it is possible that the system that we have used, where the repression loop cannot be destabilised by CRP or RNA polymerase binding, allows such a weak interaction to be detected. We also cannot exclude that the sequence between the operators in our constructs induces DNA bending or some DNA deformation that brings the two sites into close proximity. T tracts, such as those present in the sequence between operators, are characteristic of Trypanosoma minicircle kinetoplast DNA and are well-known determinants of DNA bending (24,25). This sequence might replace the requirement for HU binding, since HU is known to bend DNA (26).

Clearly, the interaction between the two dimers is not highly stable. In the absence of auxiliary proteins, cooperation between the two dimers is only seen at low temperatures on linear DNA, as revealed by quantitative analysis of repressor binding to DNA by gel retardation assay and DNase I footprinting (20). Furthermore, not only do the two gal dimers not readily interact in vitro when bound to DNA, but the gal dimer also does not show any oligomerisation in solution, even at high concentrations (27). This is an exception among the transcription factors which control regulation from multiple sites, such as the E.coli lac (28,29) and deo repressors (30,31), the E.coli arabinose AraC protein (32) and λcI repressor (33 and references therein) and the eukaryotic Sp1 (34) and p53 (35) proteins.

DNA supercoiling stabilises the weak interaction between the two dimers in vitro and in vivo: (i) in vitro transcriptional repression by GalR is only observed on supercoiled DNA (2); (ii) HU, which is supposed to facilitate interaction between the two dimers, only acts on supercoiled DNA (2–4,6); (iii) GalR–GalR DNA complexes are not formed on linear DNA, except at low temperatures, however, a non-negligible number of GalR–GalR complexes are formed on supercoiled minicircles, as revealed by electron microscopy (25% of the number of loops formed with HU); (iv) relaxation of DNA following inhibition of DNA gyrase activity by coumermycin totally relieves repression (6).

DNA superhelicity is also an essential component of DNA looping in related systems. It is not associated with HU in these systems. In vitro DNA looping of the lac repressor was first characterised on linear DNA. However, the assays were performed with high affinity consensus lacOi sites (36; M.Krämer, M.Amouyal and B.Müller-Hill, Berlin Summer School, September 1986). When the natural components were retained, DNA supercoiling was necessary in vitro (37–39) and in vivo (37). DNA supercoiling is also essential for DNA looping with AraC protein (40).

Finally, this is not the only example of an interaction that can only be characterised in vivo. For example, we have described in vivo cooperation between two lacWM sites in a previous work (9). These sites are nearly equivalent to half-sites. lac repressor binds with such a low specificity to a single site in vitro that repression of transcription is not observed when the site overlaps the promoter region in vivo. This situation has been reproduced with the lacO3–lacO3 interaction (41), which has never been demonstrated in vitro.

Along the same lines, the highly stable interaction of the lac repressor with consensus lacOi sites has been taken as a positive control for DNA looping in gal studies. However, with regard to affinities, the GalR interaction is better compared to less stable interactions, for example to the lacO1–lacO3 loop (in particular see 37) or to the lacO2–lacO3 interaction. Indeed, the affinity of GalR for its sites is only 3 × 10⁸ M^–1 (20) and it is equivalent for the two sites (27). The affinity of the lac repressor for the wild-type lacO1 site lies between 10¹⁰ and 10¹¹ M^–1 (42). The affinity of this repressor for the consensus lacOi site is 10-fold higher (43) and that for the lacO2 and lacO3 pseudo-operators is 10-fold and 100- to 1000-fold lower, respectively.

In conclusion, our observations support the hypothesis that GalR dimers are capable of non-assisted interactions in a system which allows the detection of weak protein–protein interactions in vivo. Previously we had shown that the same system could detect the cooperativity generated by weak DNA–protein interactions (9). This finding implies that repression of the E.coli gal operon may, at least partly, operate in this manner under wild-type conditions and in strains which have compensated for the deletion of HU by secondary mutations.