PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of proengLink to Publisher's site
 
Protein Eng Des Sel. 2009 November; 22(11): 673–683.
Published online 2009 September 2. doi:  10.1093/protein/gzp051
PMCID: PMC2763795

Towards evolving a better repressor

Abstract

Transcriptional regulation is an essential component of all metabolic pathways. At the most basic level, a protein binds to a particular DNA sequence (operator) on the genome and either positively or negatively alters the level of transcription. Together, the protein and its operator form an epigenetic switch that regulates gene expression. In an effort to produce a ‘better’ switch, we have discovered novel facets of the lac operon that are responsible for optimal functionality. We have uncovered a relationship between operator binding affinity and inducibility and demonstrated that the operator DNA is not a passive component of a genetic switch; it is responsible for establishing binding affinity, specificity as well as translational efficiency. In addition, an operator's directionality can indirectly affect gene expression. Unraveling the basic properties of this classical epigenetic switch demonstrates that multiple factors must be optimized in designing a better switch.

Keywords: induction, regulation, repression, transcription

Introduction

Over 40 years ago, Jacob and Monod (1961a) proposed a model for transcriptional regulation and introduced the concept of an epigenetic switch. Subsequently, it was demonstrated that this switch consists of a protein molecule (a repressor) and a particular stretch of DNA (an operator) (Gilbert and Muller-Hill, 1966; Gilbert, 1972). As a unit, the repressor and its operator are responsible for modulating the rate at which RNA polymerase can transcribe a gene. Since its conception, the lac operon has served as a model for exploring the details of transcriptional regulation. More recently, the regulatory circuitry of the lac operon has become a prominent molecular tool used to regulate gene expression in a number of systems.

One of the most widely used expression systems was created by coupling the molecular switch of the lac operon with the more efficient transcriptional machinery of the T7 bacteriophage (Studier and Moffatt, 1986). The pET expression system is now extensively used for expressing proteins in bacterial systems. The switch of the lac operon has also been shown to reversibly regulate gene expression in non-bacterial organisms (Hu and Davidson, 1987, 1988, 1990; Lee et al., 1997; Caron et al., 2005). Recently, the switch was used in stem cells to regulate gene expression during the differentiation process, as well as in mature cells (Caron et al., 2005). Using an appropriate reporter system, the kinetics of induction were shown to be ‘rapid, stable and reversible upon removal of IPTG’, illustrating that the genetic switch can regulate genes during embryonic development (Caron et al., 2005). Similarly, the lac system can be used to regulate genes in breast cancer cell lines (Lee et al., 1997). Some of the most promising work comes from the successful use of the lac system to regulate gene expression in mammals, specifically in mice (Mills, 2001). Cronin et al. (2001) created an ‘optimal inducible system’ where lac provided tight, inducible and reversible regulation of the tyrosinase gene involved in mouse pigmentation. Two lines of transgenic mice were created: one cell line expressed the lac repressor and the other expressed the tyrosinase gene under the control of the lac operator. When the transgenic mice were crossed, the double transgenic mouse contained a regulated operon.

Although many proteins and enzymes can function outside of their natural sources, their functionality is generally not optimized for use in industrial or non-native applications. The lac repressor, although widely used, is an example of a molecular tool that has not yet been optimized for its non-natural applications. The regulated range of expression obtained with the wild-type switch is well suited for its role in regulating lactose metabolism in Escherichia coli. In its natural environment, expression controlled by the switch cannot be induced unless there is a degree of leakiness allowing the natural inducer to be made (Jobe and Bourgeois, 1972). Therefore, the switch permits a basal level of background expression under repressing conditions to allow for full activation when needed. Outside of its natural environment, however, background expression of the regulated genes is not necessary since a gratuitous inducer is added directly to the system. Since leakiness of the switch can often be undesirable, we set out to improve upon the regulatory efficiency of the wild-type switch.

In order to successfully engineer more efficient transcriptional regulators, a complete understanding of each component is required. Therefore, as a first step, we have undertaken the task of dissecting the interaction between the repressor and its operator. Of utmost importance in this interaction is specificity: specificity of the repressor as well as specificity of the operator. In the absence of discrimination by the repressor for its target DNA sequence, the repressor protein would wreak havoc in the cell by interfering with alternate regulatory networks. In addition, plasticity in the recognition of repressors by the operator site would lead to non-specific regulation, and thus inefficient signaling inside the cell. For organisms to successfully regulate various cellular processes, each process must contain regulatory components specific for that pathway. In addition to specificity, the affinity between the repressor and its operator is also important. This affinity must be sufficiently high in order for the interaction to be biologically relevant at the concentrations of repressor and operator found in the cell. Too weak of an affinity would prevent association, making regulation impossible. However, the affinity must be balanced such that the interaction can be disassembled under conditions requiring transcription of the regulated gene(s).

Early efforts hypothesized that a recognition code existed with a one to one correspondence between amino acid side chains and the nucleotide bases (Seeman et al., 1976). Much effort has been spent on trying to understand specificity of protein–DNA recognition and binding affinity (Lehming et al., 1987, 1990; Sartorius et al., 1989). Despite being able to measure repression among various mutant repressors and operator sites, little work has focused on understanding the balance of affinity between the regulatory off and on states in the lac system. Therefore, designing improved transcriptional regulators is not possible without further dissection of the lac operon. The lac operon has withstood years of selective pressures and therefore, buried in its construction is a wealth of information on how to design an efficient molecular switch. Here, we have subjected both the repressor and its operator to artificial selection in an attempt to further unravel fundamental properties of a molecular switch that must be optimized when designing a more efficient switch.

Materials and methods

Creating a library of mutant repressors

A library of mutant repressor molecules was created by modifying the placI plasmid from Novagen. Two silent point mutations were introduced into the gene, creating a pair of unique restriction sites (AatII and XmaI) that flank a 76 bp fragment encoding for the key amino acids on the recognition helix (Besse et al., 1986; Lehming et al., 1987). A non-functional repressor (pBD21003) was created on the parental plasmid by introducing three point mutations, Y17P, Q18A and R22G, which allowed us to select for a gain of function and minimize the number of false positives in the selection process. To randomize the critical residues on the recognition helix, an oligonucleotide was purchase from IDT with a sequence: 5′-TTA TAC GAC GTC GCA GAG TAT GCC GGT GTC TCT NNK NNK ACC GTT TCC NNK GTG GTG AAC CAG GCC AGC CAC GTT TCT GCG AAA ACC CGG GAA AAA G-3′.

The complementary strand of this oligonucleotide was generated using a Klenow fill in reaction. In a 30 µl reaction, 1 µg of template DNA was incubated with 33 µmol of dNTPs (Invitrogen), 5 U DNA Polymerase I (Large Klenow Fragment New England Biolabs) and 30 pmol of the reverse primer (5′-CTTTTTCCCGGGTTTTCGCAG-3′ IDT). The double-stranded DNA cassette was purified using a QIAquick PCR Purification Kit (Qiagen) to remove active polymerase, salts and unincorporated dNTPs.

The recognition helix cassette was prepared for ligation by digesting 10 pmol of product with AatII and XmaI. The digest was then purified using a QIAquick PCR Purification Kit (Qiagen). The digested cassette was then ligated into the pre-digested pBD21003 plasmid, and subsequently electroporated into DH5α cells. A small fraction of the cells (5%) were plated to establish that there were 1.4 × 105 transformants. The remaining cells were used to inoculate a 25 ml cell culture. This culture was grown to saturation, and the plasmid library was purified using a QIAprep spin Miniprep Kit (Qiagen).

Creating a reporter plasmid

The reporter plasmid was constructed by placing the green fluorescent protein (GFPmut3.1) under the control of the lac operator or any operator of choice. The reporter plasmid was created by cloning a region from the pGFPmut3.1 (Clonetech) vector into the pBR322 vector between the ClaI and EcoR1 restriction sites. The fragment inserted into this region contained the lac promoter thru the end of the GFPmut3.1 gene of the pGFPmut3.1 vector. To allow for facile swapping of operator sites, a Quikchange (Stratagene) mutagenesis protocol was carried out with oligos (BD002: 5′-TTT CAC ACA GGA AAC AGC TAT GAC CAT G 3′ and BD003: 5′ TTC CAC ACA ACA TAC GAG CCG GAA G-3′) to generate a reporter plasmid without an operator (pBD1200). From this template, different lac operator variants could be cloned into a reporter by taking the aforementioned primers and adding the sequence of the right half site to the 5′ of the forward primer, and the left half site sequence to the 5′ end of the reverse primer. Each of the reporter plasmids was transformed into chemically competent DH5α E.coli cells, and the reporter cells were amplified. Each of these reporter cells was then made electrically competent.

Phenotypic screening

Fluorescent-activated cell sorting (FACS) was used to separate bacterial cells into various populations based on the intensity of their fluorescent phenotype. In short, 110 ng of the repressor library was transformed into cells containing a given reporter plasmid. A small aliquot of these cells were plated to determine the number of total transformants, whereas the remaining cells were used to inoculate a 10 ml overnight culture. The following morning, a 25 ml culture was inoculated 1/100 with the saturated overnight culture. Once these cultures reached an optical density (OD) (A600) of 0.6, the samples were transferred to a 4° shaker for 30 min. One OD of cells was then pelleted at 3000 rpm for 5 min at 4°C. The cell pellet was resuspended in sterile 5.0 ml of PBS, transferred into a sterile falcon tube and stored on ice.

Sorting was conducted at the University of Pennsylvania Flow Cytometry and Cell Sorting Facility. To establish proper population gates, positive (unrepressed) and negative (repressed) control samples were analyzed. The library samples were then sorted into each of these populations, and the cells were collected in 5 ml falcon tubes containing 1.0 ml of 50% PBS and 50% LB. Cells displaying an intermediate phenotype were also collected. Samples were then returned to storage on ice until appropriate volumes were plated on LB plates (150 × 20 mm). After colonies became visible on plates, they were chosen to inoculate overnight cultures for plasmid preparation and sequencing.

In vivo repression assay

To analyze the phenotypes of various repressor–operator combinations, an in vivo fluorescent assay was used. To quantitate the level of fluorescence and therefore indirectly measure the degree of transcription, cells were grown and analyzed in a Perkin Elmer Victor3 plate reader. In short, combinations of repressors and operators were transformed and colonies were selected for overnight culture growth. Fluorescence was monitored for cultures at all points during their growth curves and the largest dynamic range between the repressed and induced samples occurred when cultures reached stationary phase. Signals at earlier time points made it difficult to compare mutants of similar repression strength because the repressed signal could not be distinguished from background fluorescence. In general, all assays were conducted by growing each triplicate samples in either 500 µl or 1 ml of LB in a 96-well plate, aliquoting 200 µl samples at the selected time point then measuring fluorescence (495 nm excitation wavelength, 510 nm emission wavelength) and OD (A590). To normalize the signal for each sample, the signal from the blank sample was subtracted and the resulting fluorescent signals were normalized to cell OD. The normalized signals from each of the replicates were averaged, and the standard deviations were calculated (shown by error bars on plots). In addition, each repression of each repressor and operator combination was monitored using traditional B-galactosidase assays.

Reverse transcriptase–quantitative polymerase chain reaction

To analyze repression more directly, relative mRNA transcript levels were determined using reverse transcriptase–quantitative polymerase chain reaction (RT–qPCR). Each repressor–operator combination under question was transformed, and four colonies were selected to inoculate a 5 ml overnight culture with appropriate antibiotics. The following morning, 250 µl of each saturated culture was used to inoculate two fresh 25 ml cultures, one containing inducer and the other without it. At an OD A600 between 0.4 and 0.6, a 500 µl aliquot was removed and processed (Qiagen RNeasy) to isolate the RNA. Additionally, three 200 µl aliquots were taken and processed for GFP fluorescence to allow for a direct comparison of transcript levels to fluorescent signal. Upon isolation of the RNA, 10 µg of sample was treated with DNase to remove contaminating DNA (Ambion Turbo DNA-free Kit). cDNA was then generated (Invitrogen SuperScript First-Strand Synthesis System for RT–PCR) using 1.4 µg template mRNA with 75 ng of random hexamer primers. The resulting cDNA was diluted 1/10 and 1 µl was used for each triplicate reaction (Stratagene Brilliant SYBR Green QPCR Master Mix). GFP transcript levels were determined using 300 nM of forward and reverse primers (BD102F: 5′-CCA TGC CCG AAG GTT ATG TA-3′, BD103R: 5′-CGC TTC CAT CTT CAA TGT TGT-3′). To compare transcript levels from sample to sample, a housekeeping gene, Ribosomal 16S subunit, was used as a normalizer control (BD112F: 5′-GTG TTG TGA AAT GTT GGG TTA A-3′, BD113R: 5′-CCG CTG GCA ACA AAG GAT AA-3′). All samples were run and processed using thermocycler Stratagene 3000mxP.

Cloning and purification of lac DNA-binding domain mutants

The wild type and five DNA-binding domain mutants, Q18M, Y17A, Y17G, Y17R/Q18G and Y17P/Q18A/R22G, were expressed and purified identically. To allow for purification using a Nickel column, the dimeric form (residues 1–331) of each of these repressors was cloned into the pET101D vector. The pET101D vector was purchased as part of a Directional TOPO Cloning Kit (Invitrogen) that allowed for rapid (5 min) ligation of blunt ended PCR products into the primed vector. To increase the efficiency of cloning the repressor gene into the vector with the correct orientation, each repressor gene was PCR amplified (BD006: 5′-CAC CCA TAT GAA ACC AGT AAC GTT ATA CGA CG-3′, BD007: 5′-GCC CTC GAG CGC CAG CGT GGT TTT TCT TTT CAC-3′) to introduce a directionality tag at the N-terminal region of the gene. Also engineered during the PCR amplification were NdeI and XhoI restriction sites on the outer edges of the repressor sequences. The non-cleavable 6-His tag was placed at the C-terminus of the repressor genes and was shown to not interfere with protein oligomerization or DNA binding (data not shown).

The expression plasmids were transformed into the chemically competent BL21 DE3 E.coli and plated on LB plates containing Ampicillin. A colony was selected and grown in a 25 ml cultures at 37°C to saturation. The following morning, a 1-L culture was inoculated 1/100 with the saturated overnight culture. This culture was grown at 37°C until an OD (A600) of 0.4–0.6 was reached. At this time, the cultures were induced with IPTG to a final concentration of 2.5 mM. After 3 h of growth under inducing conditions, the cells were harvested at 6000 rpm (Sorvall RCB5) for 10 min at 4°C. The pellet was then resuspended in 8–10 ml of 1× lysis Buffer (50 mM NaH2PO4, 300 mM NaCl, 0.013 M BME, pH 7.4) and stored at −80°C until needed.

The frozen cells were thawed on ice, and the protease inhibitor PMSF was added to a concentration of 1.5 mM. The samples were homogenized in a cell breaker, and the lysate was then cleared by centrifugation at 15 000 rpm for 30 min at 4°C. The resulting supernatant was filtered sterilely through a 0.45 µM filter and then loaded onto a Nickel column (His GraviTrap—GE Healthcare) equilibrated with 1× lysis buffer at 4°C. The column was subsequently washed with two column volumes (~30 ml) of 1× wash buffer (50 mM NaH2PO4, 300 mM NaCl, 20 mM imidazole, 10% glycerol, 0.013 M BME, pH 7.4). To elute the His-tagged repressor protein, 3 ml of 1× elution buffer (50 mM NaH2PO4, 300 mM NaCl, 250 mM imidazole, 10% glycerol, 0.013 M BME, pH 7.4) was applied to the column. The resulting eluate was then dialyzed into 1 l of 1× GF buffer (200 mM Tris, pH 7.4, 200 mM KCl, 10 mM EDTA and 3 mM DTT) for 16 h at 4°C. After quantifying the dialyzed sample, the protein was aliquoted and frozen at −80°C until needed.

Electrophoretic mobility shift assay

Differences in electrophoritic mobility were used to determine the relative binding affinities between various repressor–operator combinations identified in the screening experiments. The DNA oligonucleotides were designed to contain clamps outside of the operator sequence to overcome annealing problems found with short nearly symmetric operator oligos.

Natural operator: BD.GS.100: TTT TCG TAT AAT GTG TGG AAT TGT GAG CGG ATA ACA ATT CTA GAC AGG AAA CG, BD.GS.101: TTT TCG TTT CCT GTC TAG AAT TGT TAT CCG CTC ACA ATT CCA CAC ATT ATA CG.

Symmetric operator: BD.GS.102: TTT TCG TAT AAT GTG TGG AAT TGT GAG CGC TCA CAA TTC TAG ACA GGA AAC G, BD.GS.103: TTT TCG TTT CCT GTC TAG AAT TGT GAG CGC TCA CAA TTC CAC ACA TTA TAC G.

Ten picomoles of the top strand were labeled with P32γ ATP using T4 polynucleotide kinase (NEB). The reaction was quenched with 20 µl of Buffer TE and 40 µl of phenol chloroform isoamyl alcohol. After extracting the aqueous phase, the probe was cleaned using a centri-spin sizing column (Princeton separations) equilibrated with TE. The labeled probe was then annealed with 1.5 M excess of the complimentary strand and diluted to 220 pM. For each binding reaction, the repressor protein was serially diluted and added to a master mix containing 1× binding buffer (10 mM Tris, pH 8.0, 250 mM KCl, 1 mM EDTA, 5% glycerol, 0.1 mg/ml BSA) and labeled probe to a final concentration of 11 pM. After incubating for 15 min, the samples were run for 2 h at 200 V at 20°C. The gels were subsequently dried at 80° for 45 min and exposed for 16–20 h on phosphor storage screens (Molecular Dynamics). The screens were scanned on a Storm 820 phosphor imager (Molecular Dynamics), and the images were analyzed with image quant software (GE Healthcare). Triplicate gels were run for each repressor–operator combination, and the resulting data were analyzed in origin to determine the dissociation constants.

Results

Plasticity of natural OR1 operator

The components of the switch that regulate the lac operon have been well characterized (Gilbert, 1972). The primary operator of the lac operon (OR1), a short stretch of DNA (~17 bp), is composed of two nearly identical half sites and is located between the end of the LacI gene and the beginning of the lacZ gene (Jacob and Monod, 1961a). The repressor is a 360 amino acid protein that has a modular structure. It contains an NH2-terminal or ‘headpiece’ domain (~60 residues) that binds specifically to operator DNA and a COOH-terminal ‘core’ domain that binds inducers. The monomeric repressor self-associates into a dimer of dimers, where a dimeric repressor molecule recognizes the operator using the classical helix-turn-helix (HTH) motif. The HTH provides the scaffold for specific side chains to recognize bases in the major groove of the operator. Structural studies (Bell and Lewis, 2000) in conjunction with genetic studies (Lehming et al., 1988; Sartorius et al., 1989, 1991) have elucidated that the most critical residues for specific operator recognition are Y17, Q18 and R22 (Fig. 1). These three amino acids protrude from the recognition helix of the repressor and form specific interactions with the bases of the operator. Although previous genetic screens were useful in demonstrating which residues and operator base pairs are important for recognition, that work was limited to analysis of non-natural, symmetric operator sites (Lehming et al., 1987). Here, we utilize a similar genetic approach but analyze a larger library of mutant repressors and the biologically relevant, natural operator.

Fig. 1
The structure of the repressor headpiece bound to the operator. The recognition helices fit into the major groove of the DNA. On the basis of a number of crystal structures, we observed three amino acid side chains that make direct interactions with bases ...

A library of ~203 or 8000 mutant repressors was created by introducing all 20 amino acids into the repressor at the three critical positions. In the past, transcriptional regulation has been evaluated by measuring β-galactosidase activity (Miller units); however, a more direct approach is to measure the concentration of the reporter protein. A plasmid was constructed such that the expression of a reporter gene, GFPmut3.1, is regulated by the lac promoter and operator (Table I). To overcome limitations in screening the repressor library on plates, functional mutants were identified by screening millions of cells and selecting molecules that block transcription of GFP using FACS. To probe specificity of the operator to discriminate various repressor molecules, mutants in the library that repress transcription of GFP were identified and their genes were isolated and sequenced. Less than 1% of the molecules in the library are capable of repressing transcription of GFP, illustrating that there are a limited number of combinations of the amino acid side chains that are capable of binding to the natural operator and blocking transcription. This finding is in agreement with previous analysis which showed that very few changes in the HTH motif are tolerated (Suckow et al., 1996).

Table I
Sequence of operators

Thirty-three different mutants were identified that block transcription of the reporter gene and therefore function as repressors (Table SI, Supplementary data are available at PEDS online). These altered repressors contain both single and double site substitutions but no functional repressors were found with changes at all three sites. Moreover, none of the functional mutants had substitutions at position 22 suggesting that the observed interaction between the arginine and the guanine at position 6 in the operator is critical. Single site substitutions account for less than half of the functional mutants; six at position 17 and four at 18. Six of the single site mutants had previously been characterized as weak operator binders in Millers herculean analysis of the repressor (Markiewicz et al., 1994). Although that genetic analysis produced enormous amounts of valuable data, the inability to produce all 20 mutants at each residue allowed us to expand upon the work and identify four novel mutants also capable of repressing the natural operator as well as probe the complementarity of mutations at more than one position. The remaining mutants we identified have substitutions at both positions 17 and 18 with a clear preference for methionine at position 18. Moreover, the presence of the Q18M substitution provides additional plasticity; creating functional repressors that allow 14 different amino acid substitutions at the first position. Since the natural operator contains two distinct half sites and the nature of the interaction in the second half site remains unclear (Bell and Lewis, 2000; Kalodimos et al., 2002), in the absence of a structure, it is difficult to determine how the methionine substitution directly affects operator binding.

These observations are consistent with previous studies that used the tight binding symmetric operator (Sartorius et al., 1991); however, it does not appear that the three mutated positions function independently of one another. We observed that the first two positions on the recognition helix function synergistically. For example, Q18L (YLR) does not bind to the natural operator (Markiewicz et al., 1994) but functionality is restored by the double mutant Y17M, Q18L (MLR). Similarly, Y17A is non-functional but the double mutant Y17A, Q18M (AMR) restores the repressed phenotype. This synergy further supports the notion that there is not a simple one-to-one correspondence for protein–DNA interactions and suggests that two neighboring residues may interact as a single unit on the protein surface. Although we observe that the 33 mutants are functional, defined as a 2-fold reduction in transcription, and decrease the rate of GFP production, they do not repress transcription equally.

Repression by all 33 mutants was measured in vivo and although the mutants are phenotypically similar to the wild-type repressor, most of the mutants do not repress transcription as effectively (Table SI, Supplementary data are available at PEDS online). There was, however, an exception: a single site mutant, Q18M, represses better than the wild-type repressor (Fig. 2A). Traditionally, repression of various repressors and operators has been determined by indirectly measuring the enzymatic activity of the reporter protein. With current technology, we could more directly measure repression by comparing transcript levels using RT–qPCR. Consistent with the observed difference in fluorescent signals, the Q18M mutant reduces the mRNA transcript more efficiently than the wild-type repressor (Fig. 2B). To verify that increased repression of the Q18M point mutant was a result of increased binding affinity to the operator and not increased stability of the protein in the cell, its binding affinity was determined in vitro. Consistent with the in vivo data, in vitro, the apparent binding affinity of the purified Q18M mutant repressor is 2.5 times greater than that of the wild-type repressor (Fig. 2C). The consistency between the data collected in vitro and in vivo suggests that the increased repression of the Q18M mutant was due to increased affinity and not increased stability of the mutant inside the cell. Although the observed affinity is lower than what has been previously published for the wild-type repressor, the discrepancy could be attributed to the changes in the repressor resulting in dimerization, the presence of a C-terminal His tag or buffer system.

Fig. 2
(A) Illustrates the normalized fluorescent signal (see Materials and methods) for samples containing GFP under control of the natural OR1 operator with the wild-type repressor and Q18M mutants. The third sample contains the ‘ideal’ symmetric ...

Many mutant repressor molecules were identified that bind to the natural operator more tightly than the wild-type repressor (Schmitz et al., 1978; Schmitz and Galas, 1980; Swint-Kruse et al., 2003); however, these amino acid substitutions are not directly involved in operator recognition. In the context of the structure, the Q18M mutant is the only tight binding repressor that interacts directly with the bases in the natural operator. Although this mutant was previously identified (M2) and shown to bind the symmetric operator with an affinity comparable to the wild-type repressor, its ability to repress the natural operator was never investigated (Lehming et al., 1987). Here, we show that altered repressors that bind specifically to the natural operator and repress better than the wild-type repressor are rare. From this analysis of ~8000 mutated repressors, only one mutant, Q18M, was identified. Therefore, we would conclude that selective pressure has established that the wild-type repressor has near optimal binding affinity to its cognate operator.

Plasticity of the symmetric operator

The operator DNA can be easily overlooked as an active participant with respect to protein–DNA interactions, but tighter repression can also be achieved by altering the operator. The wild-type repressor binds to a perfectly symmetric operator more tightly than it does to the natural sequence (Sadler et al., 1983). This tight-binding operator has a palindromic sequence that corresponds to the left half of the natural operator with the removal of the central base pair (Table I). In vivo, this higher affinity interaction translates into higher levels of repression under non-inducing conditions, essentially reducing the leakiness of the switch. To help understand what factors may have preserved a lower affinity operator site, we repeated the artificial selection experiments with the ‘ideal’ symmetric operator.

Using the repressor library, we selected mutants that bound to the symmetric operator and repressed transcription of GFP. In these selection experiments, nearly three times as many mutant repressors bound to the symmetric sequence and blocked transcription compared with the natural operator (Table SII, Supplementary data are available at PEDS online). Similar to the mutants who repress the natural operator, less than one-third of these functional mutants have single site substitutions. The greatest plasticity occurs again at the first position (Y17) and includes negatively charged (E) and positively charged amino acids (K and R). The second position is less promiscuous, yet 12 amino acid substitutions preserve the ability of the repressor to bind to the operator, whereas the third position is again invariant. These results are consistent with previous analysis of the symmetric operator (Lehming et al., 1990); however, several novel mutants were also identified (Table SII, Supplementary data are available at PEDS online).

To validate the differences in repressor recognition between the two operator sequences, repression of all of the mutants identified in the selection experiments were re-analyzed in the presence of both operators. In addition, several of the mutants were purified and their binding was analyzed in vitro. Both of these assays verified the selection results by showing that symmetry in the operator decreases specificity (Fig. 3A and B). As might be expected, all of the mutants that repress the natural operator were also found to repress the symmetric operator. However, contrary to our expectations, neither the wild-type repressor nor the Q18M mutant is the tightest binders. Several mutants were found to bind to the symmetric operator more tightly. In fact, there is no correlation in repression values when comparing the binding affinities of mutants to the each of the operators (Fig. 3C). Analyzing the specificity of repressor recognition to each of the operator sequences helps to explain why asymmetry in the operator sequence is preferred over a higher affinity interaction. By introducing slight asymmetry between each operator half site, the switch increases specificity.

Fig. 3
(A) In vivo repression assays with various repressor mutants demonstrate the differences in specificity between the natural and symmetric operators. The mutants depicted repress the symmetric operator but cannot repress transcription more than 2-fold ...

Balance of affinity and induction

Changes in both the repressor and operator that lead to a higher affinity interaction have been identified. Although higher repression reduces the leakiness of the switch by lowering the basal level of transcription under non-inducing conditions, both the repressor and operator have been preserved with less than ideal binding affinity. We hypothesized that increasing affinity between the repressor and operator may affect the repressor–operator equilibrium and alter the dynamic range of the switch. The overall effectiveness of the switch depends upon the amount of transcript produced in both the off (repressed) and on (induced) states. It is possible that if the repressor binds to the operator too tightly then the dynamic range of transcription will be compromised. To explore the relationship between binding affinity in the repressed state to the overall effectiveness of the switch, each of the repressors and operators were analyzed under inducing conditions.

The dynamic range of the genetic switch was evaluated by measuring the ratio of GFP produced in the presence and absence of inducer (Fig. 4A). For all concentrations of IPTG, the difference in fluorescence between these two states was greater for the wild-type repressor than the tight binding Q18M mutant. In both the presence and the absence of inducer, the mutant repressor binds to the natural operator more tightly. Similar results were also obtained when repression was determined by measuring the level of the GFP transcript (Fig. 4B). When the fluorescence signal regulated by the Q18M and the wild-type repressors is plotted as a function of IPTG concentration, the maximum level of transcription achieved in the presence of the Q18M mutant repressor is roughly 2.5 times lower than that observed for the wild-type repressor. This difference corresponds perfectly to the measured difference in affinity (2.5-fold) between the two repressors in the absence of inducer. Essentially, the mutant repressor binds to the natural operator more tightly in both the presence and the absence of inducer. The Q18M mutant binds to the operator and blocks transcription more efficiently than the wild-type repressor and as a consequence the conformational change resulting from inducer binding is insufficient to relieve repression. In an analogous fashion, the wild-type repressor binds to the symmetric operator in the presence and the absence of the inducer more tightly than it binds to the natural operator (Fig. 4A). Again the tighter binding is only partially relieved by induction, and the repressor binds to the symmetric operator so tightly that it cannot be fully induced even at saturating concentrations of inducer.

Fig. 4
(A) In vivo induction assays of the wild-type and Q18M mutant repressors with the natural OR1 operator and the wild-type repressor with the symmetric left (−1) operator. Under all concentrations of inducer tested (IPTG), the higher affinity repressor–operator ...

Using the thermodynamic scheme shown in Fig. 5 and assuming that (i) changes in the DNA-binding interface do not directly impact the residues involved in inducer binding and (ii) changes at residues 17, 18 and 22 do not alter non-specific DNA binding, we postulate that all of the mutations we introduced into either the repressor or the operator change the equilibrium constant KRO. Modifying the operator or the DNA binding domain alter the energetics of the protein–DNA interaction; therefore, changes in KRO are also manifested in the equilibrium constant KROI. In the context of expression levels in our in vivo system, all modifications of the operator or repressor altered the effective [R50]. Since the changes in this region of the repressor do not affect any of the other equilibrium constants, the increase in operator affinity affects repression and induction equally. Therefore, without also altering the other equilibrium properties which affect the concentration of active repressor ([Ra]), it is not possible to produce a tighter repressing switch which maintains full induction. When mutant repressors were previously identified in genetic screens, the effect of each mutation on repression was determined, but their effect on inducibility was ignored (Lehming et al., 1990). Here, the results demonstrate that a delicate balance exists between repression and induction. In order to design better switches, this balance must be preserved, and selecting switches which have been optimized for high repression alone do not produce efficient transcriptional regulators. An ideal switch must repress transcription under non-inducing conditions as well as derepress transcription when the gene products are needed. The natural switch has therefore found an ideal combination of residues in the recognition helix of the repressor and bases in the operator to efficiently balance induction and repression.

Fig. 5
Equilibrium schematic of the lac repressor. The repressor exists in two different conformations illustrated as R and R*. Mutations in the DNA-binding domain or changes in the operator result in alterations of the repressor–operator dissociation ...

Role of operator orientation

The natural operator is pseudo symmetric, possessing an approximate dyad axis about a central G:C base pair. Mutations in the left half of the operator site appeared to be more deleterious to repressor binding than the mutations in the right half of the operator; therefore, it was concluded that the repressor has a higher affinity for the left half site than it does for the right (Gilbert, 1972). In E.coli, there are two additional operators within the lac operon that are recognized by the repressor. Consistent with the difference in affinity between the two half sites of the primary operator is the sequence conservation among the three operator sites. These three sites contain high conservation in the left half site, whereas the right half site is poorly conserved. To explore the role of operator orientation on repressor binding and repression, the inverted natural operator (OR1) was incorporated into the reporter plasmid, such that the tighter binding half site was positioned more distal to the promoter (Fig. 6A). Since operator inversion included the minimal region of operator needed for binding experiments and crystallization, we anticipated that the orientation of the operator does not influence repressor binding. To test this hypothesis, we selected mutants from the repressor library capable of blocking transcription when the operator was inverted. The same mutants bound to the inverted operator and blocked transcription of GFP as were observed with the operator in the correct orientation. However, when these mutants were analyzed by in vivo repression assays, each mutant failed to repress transcription of the inverted operator as well as the natural operator (Fig. 6B). To verify that these results were not specific to the operator sequence analyzed, the OR2 operator sequence was analyzed in both orientations (Fig. 6C and D). The results were consistent with both operator sequences in the presence of each of the mutant repressors. In addition, the mRNA transcript levels were analyzed, once again showing consistent results. Since the mutant repressors cannot discriminate directionality of the operator, differences in repression must be related to the ability of RNA polymerase to transcribe its message. We suspect that when the stronger binding half site is closer to the start of transcription, the repressor can more effectively prevent polymerase from transcribing the genes of the operon. In the inverted orientation, the repressor can be displaced more easily allowing polymerase to transcribe the genes more readily. It therefore appears that the orientation of an operator is a subtle but important parameter that must be considered when creating an efficient molecular switch.

Fig. 6
(A) Sequence alignment of the OR1 and inverted OR1 operator sequence. The sequence is rotated about the central G:C base pair. (B) While transcription of GFP under non-repressing conditions is not affected by the operator orientation, when either the ...

Operator symmetry and translational efficiency

Transcription and translation of a message are usually considered to be independent events; however, we observed that symmetry in the operator alters translational efficiency. In the absence of repressor, more GFP is produced when the promoter region contains the natural operator compared with the symmetric operator, although the amounts of mRNA produced are nearly identical (Fig. 7). To examine the relationship between operator symmetry and translation of the message, a series of reporters was constructed that contained operator regions with DNA sequences that had varying degrees of symmetry. For each reporter, the fluorescent signal was measured in the absence of repressor. As shown in Fig. 7B, as the degree of symmetry increases (as determined by calculating the deltaG of a hairpin formation), the un-repressed fluorescent signal decreases. This change demonstrates that there is a clear relationship between the potential for hairpin formation and the amount of protein produced. The operator sequence resides at the beginning of the mRNA transcript, we therefore suspect that symmetry at the 5′ end of the mRNA transcript results in secondary structure formation, which interferes with the translational machinery. This is consistent with recent findings where symmetry in codons found in the 5′ region of a transcript, not the operator site, also resulted in decreased expression levels (Kudla et al., 2009).

Fig. 7
(A) Transcription of GFP and its mRNA are similar when the natural operator sequence is included in the mRNA transcript. When the operator contains the symmetric sequence, the level of transcript does not change compared with the natural sequence but ...

It was initially postulated that the repressor could alter the level of protein expression by blocking either transcription or translation (Jacob and Monod, 1961b). Subsequent analysis demonstrated that the repressor regulated the process of transcription exclusively. In this analysis, we have shown that the operator sequence, not the repressor molecule, is the component of the switch which can regulate gene expression at both the transcriptional and translational levels. It is important to note, however, that the coupling of regulation for both transcription and translation will only exist for repressible operators whose binding sites reside after the start of transcription. In contrast, most activators of transcription bind to sites upstream of the start of transcription; therefore, the sequence identity of the binding sites would have no effect on translation.

Discussion

Transcriptional regulation is complex, as it must achieve the proper balance of affinity, inducibility, specificity and translational efficiency. Altered operators and repressors exist that repress transcription more effectively than the naturally occurring system; however, optimizing repression alone has a negative effect on the induction and as a consequence the difference in transcription between the on and off (induced and repressed) states is less pronounced. Analysis of the Q18M mutant repressor and symmetric operator suggests that altering affinity by making changes in either the operator or repressor DNA-binding domain affects the affinity of the interaction whether inducer is present or absent. Since the changes we made in either the operator or the repressor recognition helix occur in a domain distinct from the inducer binding domain, we can assume that the alterations do not directly affect the energy of association between inducer and repressor (KIR). Therefore, this binding constant is nearly identical when the inducer binds to the apo-repressor (RI) or to the repressor–operator complex (ROI), and any alterations in the association constant (KRO) between the apo-repressor and operator must also affect the association between the operator and the repressor inducer complex (KRIO). Thus, optimizing repression alone, without also altering other linked equilibria, results in a switch with reduced inducibility. In other attempts to evolve the lac repressor, the same balance between inducibility and repression was exemplified. By randomly mutating the repressor and selecting for mutants with enhanced sensitivity to inducer, both Lakshmi et al. and Swint-Kruse et Al. were able to identify mutants resulting in increased inducibility. A close examination of the background expression by each of these mutants, however, demonstrates that the increased induction resulted at a decrease in repression (Swint-Kruse et al., 2003; Satya Lakshmi and Rao, 2009). To overcome the loss in repression strength, additional mutations known to increase DNA-binding affinity were incorporated to produce a mutant with the proper balance of induced and repressed expression levels (Swint-Kruse et al., 2003).

In addition to balancing repression and inducibility, an efficient switch must also balance specificity. In the lac operon, we suspect that the operator is primarily responsible for dictating the specificity of the switch. Accordingly, specificity is achieved by introducing asymmetry between each of the operator half sites since there are many more mutants that can recognize a perfectly symmetric operator than one that is pseudo symmetric. Nearly all operons in the Lac/Gal family possess pseudo symmetric operators (Weickert and Adhya, 1992), and we hypothesize that the specificity of each is inherent in the nature of the operator asymmetry. All repressor molecules that recognize palindromic operators are oligomeric and bind to the DNA as a dimer with each monomer recognizing a half site of the operator. Since symmetric operators contain two identical half sites, the repressor molecule needs only to recognize that single site, making symmetric operators less discriminating. When the symmetry is broken, the repressor must be able to recognize two different half site sequences simultaneously. Thus, the specificity of the switch is limited by how divergent the operator half sites can become while still being repressed by a homodimeric repressor containing identical DNA-binding domains.

The degree of asymmetry in the operator sequence also affects the translational efficiency of the gene(s) regulated by the switch since the operator resides at the 5′ end of the mRNA transcript. As the degree of symmetry is reduced, translational efficiency of the message increases, suggesting that pseudo symmetric operator sites are more ideal. This demonstrates that the operator portion of the genetic switch is capable of regulating both transcription and translation in the lac operon. To avoid complications of regulating translation, operator sequences with non-symmetric half sites are preferred. This may be another explanation for the conserved pseudo-symmetry displayed in the operators found in the LacI/GalR family of operons. To add to the complexity, we also observed that the orientation of pseudo-symmetric operator alters the level of transcript. Although asymmetry is an essential parameter for a switch to function efficiently, the orientation of the operator also needs to be optimized if the asymmetry generates an operator containing half sites of different repressor affinities. In the lac operon, higher repression is achieved when the stronger binding site is proximal to the promoter initiation site.

From the work presented, it is clear that selective pressure on the lac operon has resulted in a beautifully crafted switch that has been optimized to achieve a delicate balance of many factors. Accordingly, it is not possible to reduce the leakiness of the switch while maintaining full inducibility by altering the protein–DNA interaction alone. To create a better switch also requires changing either the inducer binding site or finding an inducer that can appropriately relieve the increased repression.

Funding

This work was supported by the National Institutes of Health (GM-44617).

Supplementary Material

[Supplementary Data]

Footnotes

Edited by Jacques Fastrez

References

  • Bell C.E., Lewis M. Nat. Struct. Biol. 2000;7:209–214. [PubMed]
  • Besse M., von Wilcken-Bergmann B., Muller-Hill B. EMBO J. 1986;5:1377–1381. [PubMed]
  • Caron L., Prot M., Rouleau M., Rolando M., Bost F., Binetruy B. Cell. Mol. Life Sci. 2005;62:1605–1612. [PubMed]
  • Cronin C.A., Gluba W., Scrable H. Genes Dev. 2001;15:1506–1517. [PubMed]
  • Gilbert W. Ciba Found. Symp. 1972;7:245–259. [PubMed]
  • Gilbert W., Muller-Hill B. Proc. Natl Acad. Sci. USA. 1966;56:1891–1898. [PubMed]
  • Hu M.C., Davidson N. Cell. 1987;48:555–566. [PubMed]
  • Hu M.C., Davidson N. Gene. 1988;62:301–313. [PubMed]
  • Hu M.C., Davidson N. Mol. Cell. Biol. 1990;10:6141–6151. [PMC free article] [PubMed]
  • Jacob F., Monod J. J. Mol. Biol. 1961a;3:318–356. [PubMed]
  • Jacob F., Monod J. Cold Spring Harb. Symp. Quant. Biol. 1961b;26:193–211.
  • Jobe A., Bourgeois S. J. Mol. Biol. 1972;69:397–408. [PubMed]
  • Kalodimos C.G., Bonvin A.M., Salinas R.K., Wechselberger R., Boelens R., Kaptein R. EMBO J. 2002;21:2866–2876. [PubMed]
  • Kudla G., Murray A.W., Tollervey D., Plotkin J.B. Science. 2009;324:255–258. [PubMed]
  • Lee A.V., Weng C.N., McGuire S.E., Wolf D.M., Yee D. BioTechniques. 1997;23:1062–1068. [PubMed]
  • Lehming N., Sartorius J., Niemoller M., Genenger G., von Wilcken-Bergmann B., Muller-Hill B. EMBO J. 1987;6:3145–3153. [PubMed]
  • Lehming N., Sartorius J., Oehler S., von Wilcken-Bergmann B., Muller-Hill B. Proc. Natl Acad. Sci. USA. 1988;85:7947–7951. [PubMed]
  • Lehming N., Sartorius J., Kisters-Woike B., von Wilcken-Bergmann B., Muller-Hill B. EMBO J. 1990;9:615–621. [PubMed]
  • Markiewicz P., Kleina L.G., Cruz C., Ehret S., Miller J.H. J. Mol. Biol. 1994;240:421–433. [PubMed]
  • Mills A. Genes Dev. 2001;15:1461–1467. [PubMed]
  • Sadler J.R., Sasmor H., Betz J.L. Proc. Natl Acad. Sci. USA. 1983;80:6785–6789. [PubMed]
  • Sartorius J., Lehming N., Kisters B., von Wilcken-Bergmann B., Muller-Hill B. EMBO J. 1989;8:1265–1270. [PubMed]
  • Sartorius J., Lehming N., Kisters-Woike B., von Wilcken-Bergmann B., Muller-Hill B. J. Mol. Biol. 1991;218:313–321. [PubMed]
  • Satya Lakshmi O., Rao N.M. Protein Eng. Des. Sel. 2009;22:53–58. [PubMed]
  • Schmitz A., Galas D.J. Nucleic Acids Res. 1980;8:487–506. [PMC free article] [PubMed]
  • Schmitz A., Coulondre C., Miller J.H. J. Mol. Biol. 1978;123:431–454. [PubMed]
  • Seeman N.C., Rosenberg J.M., Rich A. Proc. Natl Acad. Sci. USA. 1976;73:804–808. [PubMed]
  • Studier F.W., Moffatt B.A. J. Mol. Biol. 1986;189:113–130. [PubMed]
  • Suckow J., Markiewicz P., Kleina L.G., Miller J., Kisters-Woike B., Muller-Hill B. J. Mol. Biol. 1996;261:509–523. [PubMed]
  • Swint-Kruse L., Zhan H., Fairbanks B.M., Maheshwari A., Matthews K.S. Biochemistry. 2003;42:14004–14016. [PubMed]
  • Weickert M.J., Adhya S. J. Biol. Chem. 1992;267:15869–15874. [PubMed]

Articles from Protein Engineering, Design and Selection are provided here courtesy of Oxford University Press