Manfred Schnarr

A new LexA-based genetic system for monitoring and analyzing protein heterodimerization in Escherichia coli.

Abstract Interactions between proteins affect a wide variety of biological processes, such as signal transduction and control of gene expression. In order to facilitate the study of protein-protein interactions we have developed a new method for specifically detecting the heterodimerization of two heterologous proteins in the bacterium Escherichia coli. The assay is based on the simultaneous use of protein fusions with an altered specificity and a wild-type LexA repressor DNA-binding domain. We have tested this system with two well known eukaryotic dimerization domains (the Fos and Jun leucine zippers). The two interacting proteins were, respectively, fused to a wild-type and a mutant LexA DNA-binding domain. Their hetero-association is specifically measured by the transcriptional repression of a reporter gene (lacZ) controlled by a hybrid operator containing a wild-type half-site (CTGT) and a mutated operator half-site (CCGT). The hybrid operator/lacZ construct was integrated into the chromosome of the reporter strain (SU202) to avoid possible artefacts due to variations in plasmid copy number. This method should be particularly useful in those cases where one or both partners are also able to form homodimers, since the assay described here is sensitive only to the formation of heterodimers. Furthermore, this assay gives rise to a screenable red/white phenotype on MacConkey-lactose indicator plates, allowing for a genetic study of the specificity of the interaction.

DNA binding properties of the LexA repressor

The LexA repressor from Escherichia coli negatively regulates the transcription of about 20 different genes upon binding with variable affinity to single-, double- or even triple-operators as in the case of the recN gene. Binding of LexA to multiple operators is cooperative if the spacing between these operators is favorable. LexA recognizes DNA via its amino-terminal domain. The three-dimensional structure of this domain has been determined by NMR measurements. It contains three alpha-helices spanning residues 8-20, 28-35 and 41-54. In view of this structure, but also according to homology considerations and the unusual contact pattern with the DNA backbone, the LexA repressor is not a normal helix-turn-helix DNA binding protein like for example phage lambda repressor. LexA is at best a distant relative of this class of transcription factors and should probably be considered as a protein that contains a new DNA binding motif. A cluster of LexA mutant repressors deficient in DNA binding falling into the third helix (residues 41-54 bp) suggests that this helix is involved in DNA recognition.

Solution structure of the LexA repressor DNA binding domain determined by 1H NMR spectroscopy.

The structure of the 84 residue DNA binding domain of the Escherichia coli LexA repressor has been determined from NMR data using distance geometry and restrained molecular dynamics. The assignment of the 1H NMR spectrum of the molecule, derived from 2‐ and 3‐D homonuclear experiments, is also reported. A total of 613 non‐redundant distance restraints were used to give a final family of 28 structures. The structured region of the molecule consisted of residues 4‐69 and yielded a r.m.s. deviation from an average of 0.9 A for backbone and 1.6 A for all heavy atoms. The structure contains three regular alpha‐helices at residues 6‐21 (I), 28‐35 (II) and 41‐52 (III), and an antiparallel beta‐sheet at residues 56‐58 and 66‐68. Helices II and III form a variant helix‐turn‐helix DNA binding motif, with an unusual one residue insert at residue 38. The topology of the LexA DNA binding domain is found to be the same as for the DNA binding domains of the catabolic activator protein, human histone 5, the HNF‐3/fork head protein and the Kluyveromyces lactis heat shock transcription factor.

Activation of RecA protein: The open helix model for LexA cleavage

RecA protein is induced by the binding of DNA and ATP to become active in the hydrolysis of ATP and the cleavage of repressors. These reactions appear to depend on the structural state of the protein polymerized along the DNA, i.e. a helical coat of six RecA per turn of 95 to 100 A pitch. In support of this model of the active conformation, it was shown that high concentrations of salt also induce this helical polymerized state as well as the enzymatic activities. Here, we describe that, in vitro and with the non-hydrolyzable analogue ATP gamma S, RNA and heparin can also induce both the structural transition and the enzymatic activation of RecA to LexA cleavage in accordance with the model. RNA and heparin do not support the reaction in the presence of ATP, and they do not induce the hydrolysis of ATP either, suggesting that, in contrast to ATP gamma S, the nucleotide is not bound stably enough, and that the combined affinities of polynucleotide and ATP actually modulate the discrimination of RecA for the various possible inducers in vivo.

In vitro binding of LexA repressor to DNA: evidence for the involvement of the amino-terminal domain

Both the amino‐terminal and the carboxy‐terminal domain of the LexA repressor have been purified using the LexA protein autodigestion reaction at alkaline pH, which leads to the same specific products as the physiological RecA‐catalyzed proteolysis of repressor. We show by circular dichroism (c.d) that, upon non‐specific binding to DNA, the purified amino‐terminal domain induces a very similar if not identical conformational change of the DNA as does the entire repressor. The positive c.d. signal increases approximately 3‐fold if the DNA lattice is fully saturated with protein. Further, the amino‐terminal domain of the LexA protein binds specifically to the operator of the recA gene, producing qualitatively the same effects on the methylation pattern of the guanine bases by dimethylsulfate as the entire repressor, consisting of a methylation inhibition effect at four distal operator guanines and a slight enhancement at the central bases. The spacing between these contacts suggests that LexA does not bind to the operator along the same face of the DNA helix. As shown by c.d. studies the amino‐terminal domain harbours a substantial amount of residues in alpha‐helical conformation, a prerequisite for DNA recognition via a helix‐‐turn‐‐helix structural motif as proposed for many other regulatory proteins.

The carboxy-terminal domain of the LexA repressor oligomerises essentially as the entire protein.

The ability of the isolated carboxy‐terminal domain of the LexA repressor of Escherichia coli to form dimers and tetramers has been investigated by equilibrium ultracentrifugation. This domain, that comprises the amino acids 85–202, is readily purified after self‐cleavage of the LexA repressor at alkaline pH. It turns out that the carboxy‐terminal domain forms dimers and tetramers essentially as the entire LexA repressor. The corresponding association constants were determined after non‐linear least squares fitting of the experimental concentration distribution. A dimer association constant of K 2 = 3 × 104 M−1 and a tetramer association constant of K 4 = 2 × 104 M−1 have been determined. Similar measurements on the entire LexA repressor [(1985) Biochemistry 24, 2812–2818] gave values of K 2 = 2.1 × 104 M−1 and K 4 = 7.7 × 104 M−1. Within experimental error the dimer formation constant of the carboxy‐terminal domain may be considered to be the same as that of the entire repressor whereas the isolated domain forms tetramers slightly less efficiently. It should be stressed that the potential error in K 4 is higher than that in K 2. The overall conclusion is that the two structural domains of LexA have also well‐defined functional roles: the amino‐terminal domain interacts with DNA and the carboxy‐terminal domain is involved in dimerisation reinforcing in this way the binding of the LexA repressor to operator DNA.

Contacts between the LexA repressor--or its DNA-binding domain--and the backbone of the recA operator DNA.

Using hydroxyl radical footprinting and ethylation interference experiments, we have determined the backbone contacts made by the entire LexA repressor and its amino‐terminal fragment with the recA operator DNA. These techniques reveal essentially the same contacts between both proteins and one side of the DNA helix if one assumes that the DNA stays in the normal B‐conformation. This result is somewhat unexpected because protection of guanine bases against methylation suggested a somewhat twisted recognition surface. The backbone contacts revealed by both methods are symmetrically disposed with respect to the center of the operator, providing further evidence that the operator binds two LexA monomers. Each half‐operator contains seven interfering phosphates. These phosphates are found on both sides of the 5′‐CTGT sequence that is believed to be the principal recognition target. On the side close to the center of the operator are found two phosphates, whereas the other five are clustered on the side apart from the dyad axis. We are not aware of such an extended cluster of interfering phosphates for any other DNA‐binding protein. A quantification of the hydroxyl radical footprints allowed us to compare further the affinity of the LexA repressor for the recA operator with that of its isolated DNA binding domain. We find an only 13‐fold higher binding constant for LexA than for its amino‐terminal domain, which is in good agreement with our earlier results for the uvrA operator using a completely different binding assay.

Specific protein-DNA complexes: immunodetection of the protein component after gel electrophoresis and western blotting

A method is described to determine the presence and the relative amount of proteins within specific protein-DNA complexes. The system studied is the LexA repressor from Escherichia coli and its interaction with the operator of the caa gene encoding the bacterial toxin colicin A. After separation of the free and the complexed 32P-labeled DNA on a native polyacrylamide gel, the bound proteins are transferred on a polyvinylidine difluoride (PVDF) membrane after sodium dodecyl sulfate denaturation. Development of the protein on the membrane was achieved on reaction with an anti-LexA antibody and the use of a second anti-antibody crosslinked with alkaline phosphatase. The phosphatase activity is monitored using 5-bromo-4-chloro-3-indolyl phosphate as a substrate and 4-nitroblue tetrazolium salt. A quantitation by densitometry of both the stained protein bands on the PVDF membrane and the DNA on autoradiograms allowed us to assign the relative stoichiometry of the two different complexes formed between LexA and the caa operator. The method should allow unraveling of complicated band shift patterns arising from the presence of several binding sites for a same protein, as in our case, or from the presence of different proteins binding to a same DNA fragment.

Genetic analysis of the lexA repressor : isolation and characterization of lexA(def) mutant proteins

SummaryWe report the isolation of LexA mutant proteins with impaired repressor function. These mutant proteins were obtained by transforming a LexA-deficient recA-lacZ indicator strain with a randomly mutagenized plasmid harbouring the lexA gene and subsequent selection on MacConkey-lactose indicator plates. A total of 24 different lexA(Def) missense mutations were identified. All except three mutant proteins are produced in near-normal amounts suggesting that they are fairly resistant to intracellular proteases. All lexA(Def) missense mutations are situated within the first 67 amino acids of the amino-terminal DNA binding domain. The properties of an intragenic deletion mutant suggest that the part of the amino-terminal domain important for DNA recognition or domain folding should extent at least to amino acids 69 or 70. A recent 2D-NMR study (Lamerichs et al. 1989) has identified three a helices in the DNA binding domain of LexA. The relative orientation of two of them (helices 2 and 3) is reminiscent of, but not identical to, the canonical helix-turn-helix motif suggesting nevertheless that helix 3 might be involved in DNA recognition. The distribution of the lexA(Def) missense mutations along the first 67 amino-terminal amino acids indeed shows some clustering within helix 3, since 8 out of the 24 different missense mutations are found in this helix. However one mutation in front of helix 1 and five mutations between amino acids 61 and 67 suggest that elements other than helices 2 and 3 may be important for DNA binding.

LexA repressor induces operator-dependent DNA bending

LexA, the repressor of the SOS system in Escherichia coli induces a substantial DNA bending upon interaction with the operator of the caa gene, which codes for the bacterial toxin colicin A. Analysis by gel electrophoresis of a family of DNA fragments of identical length, but bearing the caa operator at different positions, shows that DNA bending occurs close to or within the operator sequence upon LexA binding. In contrast, the interaction of LexA with the recA operator induces no detectable bending on 5% polyacrylamide gels. This difference between the two operators is likely to be due to an intrinsic bendability of the caa operator related to thymine tracts located on both sides of the operator. Such tracts do not exist in the recA operator. The free DNA fragments harbouring the caa operator show a slight tendency to bend even in the absence of the LexA repressor. The centre of this intrinsic bend is located close to or within the caa operator.