Regulation of the Dha Operon of Lactococcus lactis

Dihydroxyacetone (Dha) kinases are a novel family of kinases with signaling and metabolic functions. Here we report the x-ray structures of the transcriptional activator DhaS and the coactivator DhaQ and characterize their function. DhaQ is a paralog of the Dha binding Dha kinase subunit; DhaS belongs to the family of TetR repressors although, unlike all known members of this family, it is a transcriptional activator. DhaQ and DhaS form a stable complex that in the presence of Dha activates transcription of the Lactococcus lactis dha operon. Dha covalently binds to DhaQ through a hemiaminal bond with a histidine and thereby induces a conformational change, which is propagated to the surface via a cantilever-like structure. DhaS binding protects an inverted repeat whose sequence is GGACACATN6ATTTGTCC and renders two GC base pairs of the operator DNA hypersensitive to DNase I cleavage. The proximal half-site of the inverted repeat partially overlaps with the predicted -35 consensus sequence of the dha promoter.

The enzymes of metabolic pathways have by and large been conserved in all kingdoms of life where they occur. In contrast, the mechanisms of pathway control are diverse. This is most obvious at the level of gene expression. The different size, structure, and sequence organization of eukaryotic and prokaryotic genomes necessitate the different control mechanisms. But even between bacteria with similar genome organizations the differences can be striking. The transcription control of dihydroxyacetone kinases is one example for such diversity, as it will be shown below.
Dihydroxyacetone (Dha) 4 kinases occur in eubacteria, animals, and plants. They can be divided into two families according to the source of high energy phosphate they utilize, ATP and phosphoenolpyruvate (PEP) (for a review see Ref. 1). The ATPdependent kinases from animals, plants, and eubacteria consist of a Dha binding and an ATP binding domain. The PEP-dependent forms consist of three protein subunits DhaK, DhaL, and DhaM (2). DhaK and DhaL are homologous to the Dha and ATP binding domains and DhaM is homologous to the IIA Mansubunits of the PEP: sugar phosphotransferase system (PTS) (3,4). DhaK is a stable homodimer of 35-kDa subunit molecular mass that binds Dha covalently by a hemiaminal linkage between the imidazole nitrogen of a histidine (His-230 in Escherichia coli) and the carbonyl carbon of Dha (5,6). DhaL contains a molecule of ADP, which in contrast to the nucleotide of the ATP-dependent kinases is not exchanged but is rephosphorylated in situ by DhaM (7). DhaM shuttles phosphate from the phosphorylcarrier protein HPr of the PTS to DhaL (2).
A BLAST analysis with DhaK and DhaL as query revealed genes for DhaK and DhaL homologs, which were associated in operons with the genes for putative transcription factors (1). These genes occur adjacent to the Dha kinase operons suggesting that they control Dha kinase expression. How this works has so far been elucidated only for E. coli (8). Here, the dha operon is controlled by DhaR, a transcription activator from the family of AAAϩ enhancer-binding proteins (EBP, Ref. 9). The Dha kinase subunits, DhaK and DhaL are corepressor and coactivator of DhaR. When Dha is present and phosphorylated, DhaL-ADP is formed, which by binding to the DhaR receiver domain stimulates DhaR activity. In the absence of Dha, the ligand-free DhaK subunit (apoDhaK) displaces DhaL from the receiver domain but unlike DhaL does not activate DhaR (8). The transcription control of the dha operon is thus coupled to the enzymatic turnover of the inducer rather than to binding of the inducer alone. This is one example to show how Dha kinase subunits sense Dha and induce their own expression. A second example is provided by the dha operon of Lactococcus lactis IL1403 that has not been characterized before. 5 This operon (Fig. 1A) for a putatively PEP-dependent Dha kinase and a glycerol-type facilitator is flanked by two genes dhaS and dhaQ.
Here we present the x-ray structures of the two subunits DhaS and DhaQ and characterize their function. DhaS belongs to the TetR family of transcription regulators (6). But unlike all known members of this family DhaS of L. lactis functions as an activator and not as a repressor of transcription. DhaQ is a paralog of the DhaK subunit, which in complex with Dha acts as the coactivator of DhaS.

EXPERIMENTAL PROCEDURES
Crystallization, Structure Determination, and Refinement of DhaS-For details of protein expression and purification see the supplemental text. Crystals of DhaSH6, a variant of DhaS with a C-terminal histidine tag, were obtained by sitting drop vapor diffusion at 20°C. The concentrated protein (20 mg/ml) was mixed with buffer containing 50 mM sodium cacodylate pH 6.0, 1.7 M ammonium sulfate, 0.015 M magnesium acetate in a 1:1 ratio. For cryoprotection, 25% glycerol was added in a serial manner. Crystals were frozen in liquid nitrogen. Mercury derivatives with different occupancies were obtained by soaking in 1 mM HgCl 2 for three months (derivative 1) and in 0.1 mM HgCl 2 for 12 h (derivative 2). To increase the phasing power the selenomethionyl L171M mutant of DhaSH6 was used as an additional derivative. The datasets for the mercury derivatives 1 and 2 and the L171M selenomethionyl derivative were collected at 110 K using a RAXIS-IV imaging plate detector mounted on a Rigaku RU-300 rotating anode x-ray generator (wavelength 1.5418 Å). High resolution diffraction data for the native crystal were collected at the BM14 beamline (wavelength 0.976 Å) of the European Synchrotron Radiation Facility (Grenoble, France) employing a MAR Mosaic225 detector (Mar Research, Hamburg, Germany). All datasets were integrated and scaled by XDS (14,15). The structure was determined by the MIRAS method. Heavy atom (mercury and selenium) sites were located by SHELXD (10) and difference Fourier methods. Phasing and solvent flipping were effected using SHARP (11) and SOLO-MON (12) resulting in a figure of merit of 0.66 to 2.31 Å resolution for the unmodified map. Automatic model building was achieved using the program ARP/WARP version 6.1.1 (13). The model was further adjusted and completed using the program O (14) version 9.0.3. Refinement was effected by the program REFMAC (15). The quality of the model was checked by PRO-CHECK (16). The results of the data collection and processing statistics are given in Table 1. The asymmetric unit of the crystal contains two DhaS monomers (one physiological dimer). The final refined coordinates consist of 178 and 179 residues in the two subunits. More than 95.7% of the residues of the final model fall within the core region and 4.3% of the residues in the allowed region of the Ramachandran plot. Atomic coordinates and structure factor amplitudes have been deposited with the RCSB, with accession codes 2IU4, and 2IU5. Comparison of the three-dimensional structures was done using the DALI server (17).
Crystallization, Structure Determination, and Refinement of apoDhaQ and the Dha⅐DhaQ Complex-Crystals of apoDhaQ were obtained after several weeks by mixing 2 l of protein (15 mg/ml) with 2 l of 0.15 M DL-malic acid pH 7.0, 20% PEG-3350 at 20°C. To prepare DhaQ for cocrystallization with Dha, 2 mM Dha was included in all the buffers used during the purification process. Crystals of the Dha⅐DhaQ complex were obtained in 0.1 M sodium acetate trihydrate, pH 4.5, 2 M ammonium sulfate. Crystals were cryoprotected by adding glycerol to 25%. The complete datasets of apoDhaQ were collected in-house (wavelength 1.5418 Å) and of the Dha⅐DhaQ complex at the Swiss Light Source (wavelength 1.2154 Å), PSI, Villigen.
The structure of apoDhaQ was solved by molecular replacement using the DhaK structure from E. coli (PDB code 1OI2) and employing the program PHASER (18). The density was further improved by CNS and an initial automated protein model was constructed into the electron density using the program ARP/WARP. The results of the data collection and processing statistics are given in Table 2. The asymmetric unit of the crystal contains two DhaQ monomers (one physiological dimer). 90.3% (90.2%) of the residues of the final apoDhaQ (Dha⅐DhaQ complex) model fall within the core region and 8.8% (8.9%) in the allowed region of the Ramachandran plot. Atomic coordinates and structure factor amplitudes have been deposited with the RCSB, with accession code 2IU6.
Determination of ␤-Galactosidase (LacZ) Activity-L. lactis IL1403 transformed with pDKQS, pDKS, and pDK ( Fig. 1B) were used for the reporter gene assay. ␤-Galactosidase activity was determined by the method of Miller (19,20). Cultures were grown in M17 medium (Difco) (supplemented with 0.5% glucose, 5 g/ml chloramphenicol, and 5 mM Dha where indi-  The same PCR product used for footprinting was also used for EMSA. 10,000 cpm of the end-labeled PCR product were incubated with 50 nM (50% saturation) and 500 nM (100% saturation) of purified DhaS without a histidine tag in 20 l of 10 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 50 mM NaCl, and 0.5 mg/ml bovine serum albumin. Where indicated 500 nM purified DhaQ and 1 M Dha were added. After 20 min of incubation at 25°C, the protein-bound DNA was separated from free DNA by native polyacrylamide gel electrophoresis (4% polyacrylamide; acrylamide-bisacrylamide 19:1, 150 V, 0.7 h) and visualized by exposure on a phosphor screen.
For primer extension total RNA was isolated from L. lactis IL1403 with the Nucleo Spin RNA II Kit (BD Biosciences). The primer P13 annealing to the codons 13-19 of dhaK was 32 P end-labeled, and the extension reaction was performed according to the primer extension system protocol of Promega. The primer extension product was analyzed on a 6% denaturing acrylamide gel.

Structure Determination and Overall Architecture of DhaS-
DhaS with a C-terminal hexahistidine tag was expressed in E. coli BL21(DE3) and purified by metal affinity and gel filtration chromatography. The final yield was 35 mg of DhaS per liter of culture (for details, see supplemental information). Details of structure determination and refinement statistics at 1.6 Å resolution are given in Table 1. The asymmetric unit contains two monomers of DhaS forming a wedge-shaped physiological dimer. Both DhaS monomers are well ordered (residues 1-179) except for the first three amino acids in the second and the last amino acid in the first monomer.
The DhaS monomer folds into nine ␣-helices (␣1-␣9) ( Fig.  2A). The DhaS monomer can be divided into an N-terminal DNA binding domain (␣1-␣3) and a core domain involved in dimerization (␣4 -␣9). The N-terminal helices ␣2 (residues 25-35) and ␣3 (residues 40 -45), which are almost perpendicular to each other, form the helix-turn-helix DNA binding motif. Helices 5 and 6 (residues 75-99) are separated by a prominent kink of ϳ117°between the fifth ␣-helical turn of helix 5 and the first of helix 6. The C-terminal helices ␣8 (residues 132 to 156) and ␣9 (residues 161-178) provide a dimerization interface of 1090 Å 2 per monomer, which covers 11% of TetR-like Activator AUGUST 11, 2006 • VOLUME 281 • NUMBER 32 the subunit surface. It contains 65% hydrophobic, 22% polar and 13% charged amino acids. As indicated by its fold, DhaS belongs to the TetR family of transcriptional regulators (21). A sequence alignment of DhaS with the Staphylococcus aureus multidrug regulator QacR (22,23) and the E. coli tetracycline repressor TetR (24,25) shows invariant residues predominantly in the DNA binding domain (␣1-␣3) whereas no sequence conservation was found in the core domain. In particular, the residues in positions critical for inducer recognition by the TetR family members (21) are not conserved in DhaS. Overlay of the DhaS and QacR monomers yields an RMS deviation of 4.0 Å for 161 out of 186 C␣, overlay with TetR a rms deviation of 4.2 Å for 134 of 198 C␣ (17). An overlay of the physiological dimers is shown in Fig. 2B. DhaS has a more open core fold and as a consequence the helix-turnhelix motifs are further apart in DhaS than in QacR. The distances between the C␣ and C␣Ј of Tyr-44 is 63 Å in the DhaS dimer, whereas it is 45 Å in QacR and 37 Å in TetR (not shown). The increased distance between the DNA binding domains of DhaS is caused by the following differences (relative to QacR) (i) different orientations of the core helices, (ii) different orientation of the DNA binding domain relative to the core helix 4, (iii) different angles between the helix axis of ␣1 (DNA binding domain) and ␣4 (core), namely Ϫ120°in DhaS but only Ϫ90°in QacR.
Structure Determination of ApoDhaQ and the Dha⅐DhaQ Complex-DhaQ with a C-terminal hexahistidine tag was expressed in E. coli BL21(DE3) and purified by metal affinity and gel filtration chromatography. The final yield was 60 mg of DhaQ per liter of culture (for details, see supplemental information). The structures of the apo-form of DhaQ and the Dha⅐DhaQ complex were solved by molecular replacement, the apo-form employing the DhaK subunit of E. coli as a search model (6), and the complex using the refined apo-model. Refinement and data collection statistics are given in Table 2. The asymmetric units of both crystal forms contain one physiological dimer. The C␣ backbone of the Dha⅐DhaQ complex is well ordered over its entire length (residues 8 -332 and 2Ј-332Ј). In the apo-form, however, residues 190 -206 and 193Ј-205Ј are disordered (Fig. 3B). The DhaQ monomer can be divided into two domains (Fig. 3A) each composed of a sixstranded sheet (yellow), which is covered by Ϫhelices (blue for the first domain, red and green for the second domain). Helices 1 of the first and 9 of the second domain form the dimer interface. An area of 2400 Å 2 (27%) of the accessible monomer surface is buried upon dimer formation. Inserted in the second domain is a cantilever like structure consisting of two short ␤-strands (red, ␤10 and ␤11) from which a loop bearing the active site His-215 is suspended (Fig. 3A).
In the structure of apoDhaQ, extra electron density in the active site could be assigned to a molecule of glycerol (Fig. 3C), which is present in the crystallization buffer. The OH groups at C1, C2, and C3 are hydrogen-bonded to the side chain of Asp-107, N⑀2 of His-54, and the main chain amide-hydrogen of Gly-51, respectively (Fig. 3C). In the Dha⅐DhaQ complex electron density can be assigned to Dha (Fig. 3D). The OH groups at  C1 and C3 are hydrogen-bonded to the side chain of Asp-107 and the main-chain amide-hydrogen of Gly-51, respectively. The geminal amino alcohol group at C2 is hydrogen bonded to N⑀2 of His-54. The carbon backbones of the non-covalently bound glycerol and the covalently bound Dha backbone as well as the hydrogen bonding residues at the topological switch points of the first domain are perfectly superimposable (Fig. 3E). Dha binding initiates a series of movements and conformational rearrangements in and around the active site (Fig. 3, B and E): (i) His-215 is pulled by 1.37 Å toward Gly-51 and by 2.33 Å toward Asp-107. (ii) The imidazole ring of His-215 is rotated around the C␤-C␥ bond, and as a consequence of this movement the hydrogen bond between N␦1 and the carboxyl O⑀1 of Glu-217 is broken. The side chain of Glu-217 is rotated away and the free space is occupied by a sulfate ion, which is present in the crystallization buffer (Fig. 3E).
(iii) Following the movement of His-215, the cantilever (red, ␤10 and ␤11) is pulled toward the N-terminal domain (blue helices in Fig. 3, A  and B). (iv) The residues 190 -206 and 193Ј-205Ј that are disordered in apoDhaQ become well ordered upon Dha binding. An extra ␤-sheet appears as a result of this ordering (␤8 and ␤9, Fig. 3, A and B). (v) The residues 134 -140 of the N-terminal domain change their conformation from aperiodic to ␣-helical (Fig. 3B). All structural alterations appear to critically depend on the covalent binding of Dha, because the non-covalently bound glycerol has no effect. Soaking with Dha makes apoDhaK crystals crack, suggesting that conformational changes are induced by the ligand and that the different space groups assumed are a consequence of the different conformations and not vice versa. For comparison, none of these differences has been noticed with E. coli DhaK (33% sequence identity). In both forms (compare PDB ID 1OI2 and 1OI3, Refs. 6 and 26) the residues corresponding to 190 -206 Notice, that the surface exposed face of helix 7 is strongly conserved in DhaQ but not in DhaK, while the entrance to the Dha (white arrow) binding site is better conserved in DhaK. The two models were built with CONSURF (28) using as input the six non-redundant DhaQ and DhaK sequences (supplemental Fig. S1), respectively, and the DhaQ structure as query. The orientation is 30°rotated around the x-axis relative to the model shown in A.
remained disordered, the Glu-232 side chain is not rotated away from His-230 and the His-230-Glu-232 pair did not change its orientation. Glu-232 has been proposed to serve as a general base, which by abstracting a proton from N␦1 of His-230 increases the nucleophilicity of N⑀3 toward the carbonyl carbon of Dha. The turning away of the carboxylate once Dha is bound may delay reprotonation of N␦1 and consequently also the (reversible) release of Dha. This delayed release may be important for maintenance of the DhaQ⅐DhaS⅐DNA ternary complex in a stable state during transcription initiation. The catalytic efficiency of the DhaK subunit, in contrast, depends on the fast exchange of substrate and product at the Dha binding site. The rigid structure of E. coli DhaK may accelerate the acid/ base-catalyzed product dissociation.

DhaS and DhaQ Are Positive Regulators of the dha Operon-
The dha operon and the dhaS gene are divergently transcribed from a 93-base pair long intergenic region (Fig. 1). To characterize the regulation of the dha operon by DhaS, low copy number plasmids containing the P dha promoter in front of the lacZ reporter gene and dhaQ and/or dhaS were constructed (Fig.  1B). L. lactis IL1403 transformed with these plasmids were grown in GM17 medium to which Dha was added as the inducer. Dha induced LacZ activity 10-fold (Fig. 4A). Deletion of either the dhaS or the dhaQ gene resulted in the complete disappearance of P dhaK -lacZ reporter gene activity, and the dha operon could no longer be induced with Dha (Fig. 4B). D,L-Glyceraldehyde induced reporter gene activity only 3-fold whereas glycerol, a structural analog of Dha, had no effect (results not shown). Given that DhaS has the fold of a well known transcription factor (TetR) (Fig. 2), and DhaQ is a paralog of the Dha binding subunit of the Dha kinase (Fig. 3) DhaS and DhaQ are activator and coactivator, respectively, of the dha operon.
Characterization of the DhaS⅐DhaQ Complex-The observation that only DhaQ and DhaS together activate transcription suggested that the two subunits might form a complex. And indeed, a stable complex between DhaS and DhaQ could be isolated when two cytoplasmic extracts containing DhaQ with a histidine tag and DhaS without a histidine tag, respectively, were mixed and purified by Ni 2ϩ -NTA chromatography (Fig.  5A, inset). The DhaS⅐DhaQ complex was then separated from excess DhaQ by gel filtration on a Superdex-75 column (Fig.  5A) and its subunit stoichiometry estimated to be 1:1 from the staining intensities of the Coomassie Blue-stained protein bands on the polyacrylamide gel (Fig. 5A, inset). The paralogous DhaK kinase subunit of L. lactis does not form a complex with DhaS pointing to the specificity of the DhaQ-DhaS interaction (Fig. 5A, inset).
The genetic analysis (above) suggested that Dha acts as the inducer of gene expression most likely through binding to one of the two subunits. To confirm this expectation, DhaQ was incubated with increasing concentrations of [ 14 C]Dha and the complex then precipitated with acetone (5). A representative binding curve is shown in Fig. 5B. Binding reaches saturation at 5 M Dha per 8 M DhaS. Although less than one this stoichiometry is compatible with one Dha binding site per DhaQ monomer shown by the x-ray structure (Fig. 3). The Dha concentration for half-maximal saturation of DhaQ (K d ) was 14 Ϯ 3 M. For comparison, DhaK of E. coli which was used as a control in these experiments has a K d of 0.8 M. DhaS did not bind [ 14 C]Dha (results not shown). These observations, namely complex formation between DhaQ and DhaS and binding of Dha to DhaQ combined with the finding that DhaS and DhaR are positive regulators of the dha operon and that Dha is the inducer together strongly suggest that DhaS utilizes as the "macromolecular inducer" the Dha-binding protein subunit DhaQ and not a low molecular weight ligand like all other known members of the TetR repressor family (21).
Characterization of the DhaS⅐DNA Operator Complex-As shown above DhaS belongs to the TetR family of transcription factors, which employ a helix-turn-helix motif for DNA binding (6). Electrophoretic mobility shift assays confirmed binding of DhaS without a histidine tag to the intergenic region between the dhaS and dhaK genes (Fig. 5C). The minimal length DNA fragment recognized by DhaS extends from bp Ϫ54 to bp Ϫ79 with respect to the start codon of the dhaKLM operon. DhaS did not bind to unrelated sequences (for instance to the promoter operator of the E. coli dha operon, results not shown) indicating that the binding reaction is specific. The DhaQ subunit added together with DhaS did, however, not produce a supershift indicating that the DhaS⅐DhaQ complex is not stable in the electrophoretic mobility shift assay.
The dha promoter/operator region was further characterized by primer extension analysis and DNase I footprinting.
The transcription start point of the dha operon was located to the adenine nucleotide at 22-bp upstream of the dhaKLM start codon (Fig. 6A). The region that is protected by DhaS without a histidine tag was identified by DNase I footprinting. It comprises 28 bp centered at 67-bp upstream of the dhaKLM start codon and 47-bp upstream of the transcription start nucleotide (Fig. 6B). Adding DhaQ with and without Dha to the incubation The structures of the low copy P dhaK -lacZ reporter plasmids used are shown in Fig. 1B. The average standard deviation for all values larger than 10 Miller units is 7%. The difference of absolute activities is caused by the fact that the activity was determined from the end-point in A and from the concentration versus time (rate) in B (19). From day to day efficiency of permeabilization and determination of cell density may further contribute to the systematic difference between experiments. mixture did not affect the pattern of the DNase I footprint (results not shown) despite the previously demonstrated roles of DhaQ and Dha for the in vivo activation of the dha promoter (Fig. 5, A and B). The 28-bp region contains an inverted repeat whose sequence is GGACACATN 6 ATTTGTCC (Fig. 6C). Two Gs between the repeats and one G in the distal repeat region become hypersensitive to DNase I-mediated cleavage. The proximal half-site of the inverted repeat partially overlaps with the predicted Ϫ35 consensus sequence of the dha promoter (TTGTCC, Fig. 6D). The two inverted repeats are at a distance of 64 Å in straight B-DNA (measured between the hypersensitive G in the proximal and the mirror G in the distal repeat). This distance is of the same order as the 63 Å distance between the DNA binding helices of the DhaS dimer (Fig. 2). A rigid body docking of DhaS with straight B-DNA is however not possible because the DhaS core protrudes into the open space between the DNA-binding domains. To fit the recognition helices into the major grooves, the DNA-binding domains of DhaS must either rotate away from the core domain, or the target DNA must bend. A peak of DNA curvature is predicted for the proximal repeat and a minimum of curvature for the Ϫ10 promoter sequence by BEND, a program to calculate the macroscopic curvature of DNA (27). This DhaS-protected sequence is strongly conserved in four out of six dhaKLM-dhaS intergenic regions from dha operons encoding non-redundant DhaS, DhaQ and Dha kinase subunits (Ͻ90% pairwise sequence identity). All four are of lactococcal and streptococcal origin suggesting a common mechanism of dha operon control. The remaining two putative promoter regions of Bacillus halodurans and Bacillus cereus, are similar to each other but unrelated to the lactococcal sequences (not shown).
The DhaQ-DhaS Binding Interface-A comparison of the conserved amino acid sequence patterns of six non-redundant DhaQs with the corresponding six DhaKs was used to identify the putative protein-protein binding surface. DhaQ and DhaK interact with different protein subunits; DhaQ with DhaS, and DhaK with the catalytic phosphotransferase subunit DhaL (Fig. 1A). Assuming that functionally important interfaces are better conserved than the average surface, the amino acid sequences of the DhaQ and DhaK subunits were aligned separately (supplemental Fig. S1). The similarity patterns were then projected onto the experimental DhaQ structure for comparison of the structures by using the program CONSURF (28) (Fig. 3, F and G). In DhaQ helix 7 is strongly conserved mainly caused by the conservation of five surface exposed, ␣-helically phased residues, which are not conserved in DhaK. We therefore propose that helix 7 participates in DhaS binding and that the conformational change triggered by Dha binding to His-215 is propagated through the cantilever to helix 7 and hence to DhaS. A second difference is seen between the surfaces around the Dha binding sites (yellow), which is better conserved in the catalytic DhaK subunit than in DhaQ. It is to this area that the DhaL-ATP subunit must dock for the transfer of phosphate from ATP to Dha (2). Overall, the amino acid sequences of the N-terminal ␣/␤ fold (residues 1-182) is better conserved in the DhaKs (61 Ϯ 8% identity) than in the DhaQs (44 Ϯ 7%). For the C-terminal domain the difference of conservation is smaller, namely 56 Ϯ 10% for DhaK and 47 Ϯ 6% for DhaQ.

CONCLUSION
The transcription regulators of the TetR family whose functions have been described so far (85 out of 2350) are repressors, which are inactivated by small ligands such as antibiotics, drugs, catabolites, osmoprotectants, and quorum-sensing autoinducers (21). DhaS which undoubtedly also belongs to this family, is different. It is an activator of transcription that is activated by a protein ligand, DhaQ. Like many other transcription activators, DhaS binds to an operator sequence upstream of and only partially overlapping with the Ϫ35 promoter consensus sequence (29), whereas the QacR and TetR repressors bind to inverted repeats overlapping with or downstream of the Ϫ10 region of the promoter (30,31). Parenthetically, the activating function of DhaQ must not be confused with the function of "reverse" TetR mutants that bind DNA only in the pres-ence of a ligand (32), unlike wild-type TetR that dissociates from DNA in the presence of the inducer. Electrophoretic mobility shift as well as DNase footprinting indicate that binding of DhaS to the operator DNA is not influenced by the coactivator DhaQ and the inducer Dha. Similar behavior has been observed with PobR the transcription activator for p-hydroxybenzoate hydroxylase (33). The DhaQ⅐Dha complex may induce a conformational change in DhaS, leading to enhanced RNA polymerase binding. It is not known, whether the DhaS⅐DhaQ complex activates transcription by distorting the DNA structure or by allosteric interaction with the sigma RNA polymerase.
A BLAST search of the microbial genomes for DhaS homologs produced eleven dhaS genes that were genetically linked to dhaQ genes and operons for Dha kinases. All of them occur in genomes from Bacillacales and Lactobacillacales. The six DhaSs with non-redundant amino acid sequences (Ͻ90% sequence identity; Swiss-Prot IDs Q9CIV9, 8P228, Q8E3R3, Q9K7G7, Q73CL2, Q88ZX8) display pairwise sequence identities between 22 and 45% over the full-length. The identities are between 35 and 65% for the DNA binding domain (␣1-␣3) and helix ␣4, between 25 and 30% for the kinked helix 5 (␣5) whereas the dimerization helices ␣6 -␣8 are not conserved. The second most closely related group of DhaS homologs (for instance Q8Y9P4 and Q893H0) is encoded by genes associated with putative myosin-cross-reactive antigen, short chain oxidoreductases, and hypothetical proteins. One representative with an open structure like DhaS is TM1030 of Thermotoga maritima (Fig. 2C). However, none of these proteins has been functionally characterized and it is thus not known, whether they are activators like DhaS or repressors like TetR.
In conclusion, control of the dha operon in L. lactis and E. coli occur by different mechanisms. In E. coli, the DhaK and DhaL subunits have a dual function. They catalyze phosphorylation of Dha and at the same time act as mutually antagonistic corepressor and coactivator of an enhancer binding protein (8). In L. lactis, DhaQ, a paralog of the catalytic DhaK subunit, acts as the inducer binding coactivator of a transcription activator from the TetR family. L. lactis and E. coli respond to the same inducer, Dha, but they employ different mechanisms to regulate dha gene expression.