Structural and Functional Characterization of Escherichia coli Toxin-Antitoxin Complex DinJ-YafQ*

Background: The type II TA dinJ-yafQ module autoregulates bacterial growth in response to environmental stimuli. Results: The crystal structures of the DinJ-YafQ complex and free YafQ unveiled their structural details. Conclusion: The DinJ-YafQ complex as a transcription repressor interacts with a defined region in its operator via the RHH domain of DinJ. Significance: This study defines YafQ as a ribosome-dependent ribonuclease in vivo. Toxin YafQ functions as a ribonuclease in the dinJ-yafQ toxin-antitoxin system of Escherichia coli. Antitoxin DinJ neutralizes YafQ-mediated toxicity by forming a stable protein complex. Here, crystal structures of the (DinJ)2-(YafQ)2 complex and the isolated YafQ toxin have been determined. The structure of the heterotetrameric complex (DinJ)2-(YafQ)2 revealed that the N-terminal region of DinJ folds into a ribbon-helix-helix motif and dimerizes for DNA recognition, and the C-terminal portion of each DinJ exclusively wraps around a YafQ molecule. Upon incorporation into the heterotetrameric complex, a conformational change of YafQ in close proximity to the catalytic site of the typical microbial ribonuclease fold was observed and validated. Mutagenesis experiments revealed that a DinJ mutant restored YafQ RNase activity in a tetramer complex in vitro but not in vivo. An electrophoretic mobility shift assay showed that one of the palindromic sequences present in the upstream intergenic region of DinJ served as a binding sequences for both the DinJ-YafQ complex and the antitoxin DinJ alone. Based on structure-guided and site-directed mutagenesis of DinJ-YafQ, we showed that two pairs of amino acids in DinJ were important for DNA binding; the R8A and K16A substitutions and the S31A and R35A substitutions in DinJ abolished the DNA binding ability of the DinJ-YafQ complex.

Toxin YafQ functions as a ribonuclease in the dinJ-yafQ toxin-antitoxin system of Escherichia coli. Antitoxin DinJ neutralizes YafQ-mediated toxicity by forming a stable protein complex. Here, crystal structures of the (DinJ) 2 -(YafQ) 2 complex and the isolated YafQ toxin have been determined. The structure of the heterotetrameric complex (DinJ) 2 -(YafQ) 2 revealed that the N-terminal region of DinJ folds into a ribbonhelix-helix motif and dimerizes for DNA recognition, and the C-terminal portion of each DinJ exclusively wraps around a YafQ molecule. Upon incorporation into the heterotetrameric complex, a conformational change of YafQ in close proximity to the catalytic site of the typical microbial ribonuclease fold was observed and validated. Mutagenesis experiments revealed that a DinJ mutant restored YafQ RNase activity in a tetramer complex in vitro but not in vivo. An electrophoretic mobility shift assay showed that one of the palindromic sequences present in the upstream intergenic region of DinJ served as a binding sequences for both the DinJ-YafQ complex and the antitoxin DinJ alone. Based on structure-guided and site-directed mutagenesis of DinJ-YafQ, we showed that two pairs of amino acids in DinJ were important for DNA binding; the R8A and K16A substitutions and the S31A and R35A substitutions in DinJ abolished the DNA binding ability of the DinJ-YafQ complex.
Toxin-antitoxin (TA) 3 systems are widespread in bacteria and archaea (1)(2)(3) and play crucial roles in the regulation of cell growth and cell death initiated by the stress response, the SOS response, biofilm formation, and multidrug resistance (4 -8).
Located on plasmids or chromosomes, TA systems typically comprise two genes organized in an operon that codes both for a stable toxin and a labile antitoxin. Under normal cellular conditions, antitoxins typically interact with their cognate toxins to inhibit toxicity, allowing normal cell growth. However, under stress conditions, the expression of TA systems is decreased, leading to an imbalance in the amount of antitoxins and toxins present in the cell. Because toxins are more stable than antitoxins, toxins present in stressed cells are released, causing growth inhibition or even cell death.
Toxins are characterized as proteins, whereas antitoxins are either proteins or small noncoding RNAs. Currently, TA systems are classified into five classes (types I-V) according to the nature and mode of action of the antitoxin (9). In type I and III TA modules, the antitoxins are small noncoding RNAs (9). In type I TA systems, antisense RNAs act as antitoxins and inhibit translation of their cognate toxins by binding to the toxin-encoding mRNAs (10,11). For example, the symR/symE module of Escherichia coli is a type I TA system (12). However, in type III TA systems, RNA antitoxins bind to the toxins directly. By forming a complex, RNA antitoxins neutralize a type III TA toxicity (10,13,14). The only reported example of a type III TA system is the toxI/toxN module from Pectobacterium carotovoum (9,13,14). In type II TA systems, a protein antitoxin blocks the toxicity of a toxin by forming a stable complex (10). In type IV TA systems, the protein antitoxin cannot form a complex with its cognate toxin but acts as an antagonist for its toxicity (e.g. the yeeU/yeeV module from E. coli) (15). In type V TA systems, a protein antitoxin inhibits its cognate toxin by specifically cleaving its mRNA (e.g. the ghoS/ ghoT module from E. coli) (16).
Although toxins generally exert their functions in crucial cellular processes, such as translation, DNA synthesis, cytoskeleton formation, membrane integrity, and cell wall biosynthesis, most of the characterized toxins, such as RelE, are endoribonucleases and inhibit translation by cleaving mRNAs with different specificity (10,18). According to sequence similarities, the toxins YafQ, YoeB, HigB, and YhaV are classified within the RelE family (27,36,37). RelE, which contains a microbial RNase fold, is a ribosome-dependent RNase cleaving the mRNA codon positioned at the A-site in the ribosome, between the second and third nucleotides (38). However, antitoxin RelB wraps around RelE in the RelBE complex, thereby preventing entry of the toxin into the ribosome A-site and abolishing its toxicity (19).
In the E. coli dinJ-yafQ TA system, YafQ toxin, which also contains a microbial RNase fold, is an endoribonuclease that associates with the ribosome, and its overproduction causes growth inhibition or even cell death due to its RNA cleavage (7,22). In vivo, YafQ selectively cleaves mRNA codons positioned in the A-site of the ribosome (7). Unlike RelE, YafQ exhibited robust ribosome-independent ribonuclease activity in vitro. Under normal growth conditions, antitoxin DinJ forms a stable complex with toxin YafQ, sequestering its incorporation into ribosomes and neutralizing its toxicity. However, stress-induced ATP-dependent proteases preferentially eliminate unstable DinJ, resulting in YafQ release and incorporation into ribosomes. DinJ and the DinJ-YafQ complex can autoregulate their expression through binding the upstream sequence of the dinJ-yafQ module (7,10,22). There are three imperfect palindromic sequences (palindromes pal I, pal I-II, and pal II), spanning the region Ϫ62 to Ϫ12 with respect to the translation start site of the dinJ-yafQ module of E. coli. They were previously reported as putative dinJ-yafQ operator regions (22). Palindromes pal I and II are juxtaposed with one base pair overlapping, whereas pal I-II comprises the 3Ј region of pal I and the 5Ј region of pal II. Pal II was previously shown to harbor a putative LexA box (7). Recent studies showed that only palindromes pal I and I-II are the response sites for autoregulation (22).
To date, the TA crystal structures of YoeB, YoeB-YefM, and MazF-MazE from E. coli and the RelE-RelB complex from E. coli and Pyrocccus horikoshii have been solved (17,21,25,30). However, the structure of DinJ-YafQ has not yet been determined for any bacteria or archaea. In this study, we present the crystal structure of the DinJ-YafQ complex from E. coli at a resolution of 2.1 Å. The crystal structure revealed that DinJ consists of three domains: an N-terminal domain, a middle loop (linker), and a C-terminal domain. The N-terminal domain is mainly involved in homodimerization and also participates in DNA binding. The C-terminal domain of DinJ wraps around the globe-shaped YafQ. DinJ inhibits the toxicity of YafQ with the help of its intermediate loop covering the predicted active center of YafQ. Compared with the crystal structure of free YafQ, a conformational change in the positioning of the YafQ loop (␣2-␤3) in DinJ-YafQ in the vicinity of catalytic site was identified, which disrupts YafQ ribonucleolytic activity. By an electrophoretic mobility shift assay (EMSA), we also identified the shortest DNA fragment able to bind to the DinJ-YafQ complex. Based on the structure of DinJ-YafQ, we used site-directed mutagenesis of DinJ to determine which amino acids are important for DNA binding. The R8A and K16A substitutions and the S31A and R35A substitutions of the predicted binding site residues of DinJ abolished DNA binding, as shown by an EMSA.
The expression strain was grown in shaker flasks to approximately A 600 0.6 at 37°C and induced at 30°C for 6 h with 0.2 mM isopropyl ␤-D-thiogalactopyranoside. After induction, cells were harvested, and pellets were flash-frozen. For performing protein purification, cell pellets were suspended in 50 mM Tris-HCl (pH 8.0), 20% (w/v) sucrose, 350 mM NaCl, 20 mM imidazole, 0.1% Tween 20, 1 mM PMSF, 1 mM ␤-mercaptoethanol, and 10 g/ml DNase. Cell suspension was disrupted by a high-pressure homogenizer. Cell homogenate was centrifuged. Supernatant was subjected to nickel-nitrilotriacetic acid bead column purification. His-tagged protein or complex was eluted in 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, 250 mM imidazole, and 1 mM ␤-mercaptoethanol. Smt3 fusion proteins were cleaved by recombinant Ulp1 overnight at 4°C and purified with HiTrap heparin HP and Superdex-75 16/60 chromatography (GE Healthcare). For purifying YafQ, mutants YafQ(H23Q) and YafQ(K48A), and the N-terminal hexa-His and Smt3 fusion of YafQ, mutants YafQ(H23Q) and YafQ(K48A), in complex with DinJ, were initially purified. They were then subjected to a series of denaturation/renaturation steps, according to Ref. 7. Refolded Smt3-fused proteins were treated by Ulp1 digestion, and finally tag-removed YafQ and mutants were further purified with a Hitrap SP column and gel filtration chromatography (GE Healthcare). Protein fractions after chromatography were analyzed by SDS-PAGE.
Dimethyl Pimelimidate Cross-linking of Protein DinJ-Dimethyl pimelimidate (Sigma) was added to the reactions at a final concentration of 10 mM. Protein DinJ was diluted to a final concentration of 0.1 and 0.2 mM in buffer (50 mM Hepes-KOH, pH 8.5, 50 mM NaCl, 5 mM MgCl 2 ). The reactions were incubated for 1 h at 37°C. The reactions were then stopped by the addition of a final concentration of 25 mM Tris-HCl, pH 6.8, followed by adding 2ϫ SDS loading buffer. The samples were analyzed by SDS-PAGE.
Using a sitting drop-vapor diffusion method, crystals of YafQ, YafQ(H87Q), and Se-Met YafQ(H87Q) were obtained at 20°C in crystallization solution containing 0.2 M (NH 4 ) 2 SO 4 and 30% PEG 8000. Adding 1 mM Pr(III) acetate hydrate to crystallization solution was optimal for the better crystal formation of Se-Met YafQ(H87Q). Sixteen percent (v/v) of glycerol was used for cryoprotection.
Diffraction data were collected at beamline BL17U1 of the Shanghai Synchrotron Radiation Facility and beamline 1W2B of the Beijing Synchrotron Radiation Facility. Data were processed with the HKL2000 program package (40). The structure of the Se-Met DinJ-YafQ complex and Se-Met YafQ(H87Q) were solved by the single wavelength anomalous diffraction method. Selenium sites were identified, and the initial phases were calculated using the programs ShelxD and ShelxE (41), respectively. Density modification with noncrystallographic symmetry averaging was performed using the program Resolve (42) to improve the initial phases, producing experimental electron density maps of excellent quality. Automated model building was performed with the program ARP/warp (43). DinJ-YafQ native structure (wild type) and YafQ(H87Q) mutant structure were determined by molecular replacement with Phaser (44), using the structure of Se-Met DinJ-YafQ and Se-Met YafQ(H87Q), respectively, as the search model. All structures above were refined with the program Phenix.refine (45) and manually corrected in Coot (46). The qualities of the final models were checked with the program MolProbity (47). The crystal of YafQ belonged to space group C2221 with unit-cell parameters a ϭ 41 A summary of data collection and final refinement statistics is given in Table 2. The program PyMOL (Schroedinger, LLC, New York) was used to prepare all structural figures.
EMSA-Each DNA duplex for EMSA was created by annealing two complementary oligonucleotides. The 369-bp DNA fragment corresponding to the hypothetical dinJ-yafQ promoter/operator region was obtained by PCR amplification. The EMSA reaction (10 l) was carried out at room temperature, by mixing 5 pmol of a DNA duplex and a defined concentration of different protein in a binding buffer (5% glycerol, 10 mM Tris-HCl (pH 8.0), 5 mM MgCl 2 , 100 mM NaCl, 10 mM ␤-mercaptoethanol). After incubation for 30 min, the resultant was electrophoresed on a 6% polyacrylamide gel with 0.25ϫ TBE, and the gel was visualized by ethidium bromide staining.
Total RNA Extraction and RNA Cleavage Analysis in Vitro-Total cellular RNA was isolated from the E. coli DH5␣ strain, using the hot phenol extraction method (48). For RNA cleavage, total RNA was utilized for YafQ-mediated cleavage in vitro. The cleavage reaction (10 l), containing 5 g of total RNA and 0.5 g of the YafQ or mutant protein, was incubated for 15 min at 37°C in the reaction buffer (10 mM Tris-Cl, pH 8.0). The reaction was quenched by adding an equal volume of 2ϫ loading dye buffer (New England Biolabs) and heated at 70°C 3 min, prior to running on a 1% agarose gel. The agarose gel was visualized by ethidium bromide staining.
Crystal Structure of the DinJ-YafQ Complex-A DinJ-YafQ crystal was obtained by the hanging drop-vapor diffusion method at 20°C. The structure of the YafQ protein complexed with DinJ was determined to a resolution of 2.1 Å by the single wavelength anomalous diffraction method, using a crystal of the selenomethionine-substituted DinJ-YafQ derivative ( Table 2). The structure comprised eight copies of DinJ-YafQ heterodimers in the crystallographic asymmetric unit, consisting of four (DinJ) 2 -(YafQ) 2 heterotetramers related to each other by noncrystallographic symmetries (Fig. 2A). The (DinJ) 2 -(YafQ) 2 heterotetrameric form observed here was consistent with the oligomeric state described above, as determined by analytical gel filtration chromatography in solution (Fig. 1A). Each (DinJ) 2 -(YafQ) 2 tetramer constitutes three globular domains as a triangular assembly: a DinJ homodimerization domain and two heterodimerization domains, each comprising a whole YafQ molecule and a DinJ C-terminal domain ( Fig. 2A). In the resulting heterotetramer, the triangle-shaped structure with a YafQ-(DinJ) 2 -YafQ architecture of (DinJ) 2 -(YafQ) 2 occupies an area of approximately 105 ϫ 360 Å. The final structure covers most of YafQ (residues 3-92 of 92) and DinJ (residues 3-86 of 86).
In the crystal structure of the complex, the YafQ monomer forms a compact globular structure with an ␣/␤/␣-fold. This protein consists of a central four-stranded ␤-sheet (␤1-␤4) core surrounded by four ␣-helices, denoted ␣1-␣4. Together with the ␤-sheet core, helix ␣4 and the loop connecting helix ␣3 and strand ␤2 form a cleft. Notably, one sulfate ion (SO 4 -1) is embedded in the cleft (Fig. 2B) coordinated by residue His-50, Asp-61, His-63, and His-87 of YafQ via hydrogen bonds. This cleft possibly represents the catalytic site for YafQ. The catalytic site of the toxin YafQ mostly harbors aromatic or hydrophobic residues that are thought to accommodate the substrate mRNA. In contrast, the DinJ monomer resumes a more elongated shape in the complex crystal structure, with the topology ␤1, ␣1, ␣2, ␣3, ␤2, and ␣4. This elongated structure can be divided into three portions: an N-terminal domain (␤1-␣1-␣2), a middle-coiled linker, and an extended C-terminal domain (␣3-␤2-␣4). The N-terminal domain of DinJ presents a typical ribbon-helix-helix (RHH) motif, which dimerizes with the RHH motif of an adjacent DinJЈ to form a putative DNA-binding domain in the structure (Fig. 2A). The ␤1 strands of antitoxin DinJ and the adjacent DinJЈ form an intermolecular short antiparallel ␤-sheet that contributes to stabilization of the heterotetramer (DinJ) 2 -(YafQ) 2 . Three sulfate ions (SO 4 -2, SO 4 -3, and SO 4 -4) are located on this combinatory ␤-sheet surface over the dimerization domain, possibly representing the location for cognate DNA binding (Fig. 2A). The extended C-terminal domain (residues 50 -79; ␣3-␤2-␣4) entirely wraps around YafQ, with strand ␤2 of DinJ forming an antiparallel sheet with ␤1 of the toxin YafQ. Helices ␣3 and ␣4 of DinJ clip YafQ by a hydrophobic interaction. The middle-coiled linker of DinJ covers the proposed catalytic cleft of YafQ. Residues Leu-47, Pro-48, and Leu-51 on the linker of DinJ also form hydrophobic interactions with the N-terminal domain of DinJ. Notably, oneinteraction formed between the side chain of residue Phe-49 on the linker and the side chain of His-50 in YafQ may be interesting and may contribute to blocking potential RNA substrate access to the catalytic cavity ( Fig. 2B; see  below).
Crystal Structure of Isolated YafQ and YafQ(H87Q) Mutant-To further assess the structural mechanism behind ribonuclease catalysis and toxicity, the structures of WT YafQ of E. coli and its mutant YafQ(H87Q) were determined (for related structural details, see Table 2). The crystal structures of WT YafQ and mutant YafQ(H87Q) in isolation were almost identical except for the mutated residue (Fig. 2C). When overlaid onto the crystal structure of the DinJ-YafQ complex, the crystal structure of WT YafQ superimposed well, except at one site (see below), the loop connecting helix ␣3 and strand ␤2 (loop ␣3-␤2, a switch loop), which appears to rearrange into a closed and compressed configuration in isolated YafQ. Loop ␣3-␤2 in the isolated YafQ swings toward the ␤-sheet core, narrowing the passage that would accommodate the interaction of DinJ helix ␣3 in the TA complex. The positioning of Trp-56 in the ␣3-␤2 loop alters most dramatically.
Possible Functional Site of YafQ and YafQ Ribonuclease Activity in Vitro-As a RelE family member, the structure of YafQ contains a microbial RNase fold that more closely resembles RNase T1 from Aspergillus oryzae (Fig. 2D) than the RelE and YoeB structures. We superimposed the structure of RNase T1 (Protein Data Bank entry 1B2M) from A. oryzae in complex with GpU with the isolated YafQ structure (Fig. 2D). We noticed that, despite a lack of sequence homology between YafQ and RNase T1, key residues at the active site were highly conserved. In RNase T1, the active site is composed of residues Tyr-38, His-40, Glu-58, Arg-77, His-92, and Phe-100 (Fig. 2D) (49). After superimposing the YafQ structure, His-50, Asp-61, His-87, and Phe-91 of YafQ overlapped with His-40, Glu-58, His-92, and Phe-100 of RNase T1, respectively. However, residues Tyr-38 and Arg-77 of RNase T1 were substituted by Lys-48 and Ile-69, respectively, in the YafQ structure (Fig. 2D). His-63 in the neighboring ␤ strand of YafQ might play a role similar to that of Arg-77 of RNase T1, although when their structures are superimposed, Ile-69 of YafQ is equivalent to Arg-77 of RNase T1. Catalysis of YafQ may proceed via a mechanism proposed for RNase T1, whereby a protonated His-87 and an unprotonated Asp-61 in YafQ probably constitute the catalytic acid-base pair. However, mutation analysis indicated that the identity of residue Asp-61 in YafQ was not essential for cell viability. It is possible that His-50 of YafQ, which is replaced by Glu-46 in YoeB, matches with His-87 as the catalytic acidbase pair. Alternatively, as proposed for YoeB, it is likely that base A1493 of 16 S rRNA might act as a general base in the absence of Asp-61 in vivo (50) . Both YafQ and YoeB, retaining a complete set of catalytic residues for the RNase fold, are capable of cleaving certain RNAs in vitro, whereas RelE is not. However, in vivo, the role of ribosome recruiting in a RelE family toxin at its A site may therefore be to stabilize mRNA in a conformation that facilitates cleavage (38).
In YafQ crystal structures, an embedded sulfate ion (SO 4 -A in isolated YafQ, or SO 4 -1 in complex) is coordinated by residues His-50, Asp-61, His-63, and His-87 in the putative catalytic cleft. This sulfate is believed to mimic the scissile phosphate of substrate RNA. Modeling the nucleotide GpU in the  JULY 25, 2014 • VOLUME 289 • NUMBER 30

Structure and Function of Toxin-Antitoxin Complex DinJ-YafQ
RNase T1 structure into the active site of isolated YafQ shows that the phosphate of 3Ј-GMP overlaps well with SO 4 -A. Additionally, a second sulfate, SO 4 -B, in the isolated YafQ structure is 7.1 Å away from SO 4 -A. The proximity of these two sulfates might imply how an RNA substrate is situated over the YafQ active site in the process of cleavage.
To test YafQ ribonuclease activity in vitro, we assayed recombinant YafQ using total cellular RNAs extracted from E. coli strain DH5␣ as substrate. Wild-type YafQ was purified from the DinJ-YafQ complex through a series of denaturation/ renaturation steps. The results showed that WT YafQ efficiently cleaved DH5␣ total RNA in a salt-sensitive manner, but the DinJ-YafQ complex and mutant YafQ(H87Q) failed to cleave DH5␣ total RNA. Mutant YafQ(H87Q) retained the interaction with DinJ to form the heterotetramer. We further characterized mutants YafQ(K48A) (equivalent to Lys-44 of YoeB) and YafQ(H23Q). Based on structural analysis, residue Lys-48 of YafQ was found to be in close proximity to the catalytic site and 3 Å away from the second sulfate SO 4 -B in the YafQ crystal structure (Fig. 2D), indicating that this residue is likely to be important in catalysis or RNA binding. Residue His-23 may also play a role in catalysis in the YafQ dimeric form, as observed with the YafQ crystal. In the free YafQ structure, residue His-23 was close to the active site of a neighboring YafQ in the crystal packing. However, an in vitro assay for ribonuclease activity showed that in both the K48A and H23Q mutants, a moderate decrease in YafQ ribonuclease activity occurred (Fig. 3A).
Hydrophobic Interaction between DinJ Helix ␣3 and YafQ Contributes to DinJ Antidote Suppressing YafQ Ribonuclease Activity-In the DinJ-YafQ complex structure, YafQ toxicity was blocked due to its interaction with antidote DinJ, and YafQ toxicity is known to be caused by its RNase activity. The YafQ RNase catalytic region is covered by the loop linker and helix ␣3 of DinJ (Figs. 2B and 3B). The hydrophobic interactions between DinJ helix ␣3 and YafQ expel loop ␣3-␤2 of YafQ from the ␤-sheet core, leading to a local conformational change in YafQ, compared with the structure of free YafQ. The positioning of Trp-56 in the ␣3-␤2 loop alters most dramatically. This change in YafQ might interfere with its ribonuclease activity and toxicity. Next, we investigated whether any perturbations of these interacting interfaces could release YafQ RNase activ- ity. Two DinJ mutants were constructed (F49A and I59N/ I62N), and their interaction with YafQ was analyzed. The side chain of residue Phe-49 in the loop linker of DinJ interacts with His-50 of YafQ via ainteraction (Fig. 2B). Residues Ile-59 and Ile-62 located in helix 3 of DinJ directly interact with YafQ by a hydrophobic interaction (Fig. 3C). The growth of both mutant strains was unaffected, indicating that the DinJ variants neutralized YafQ toxicity similarly to WT in vivo. It is probable that, in vivo, YafQ was still capable of binding to these DinJ mutants but could not incorporate ribosomes and therefore lost its ribosome-dependent RNase activity. Indeed, in vitro, both the DinJ(F49A)-YafQ and DinJ(I59N/I62N)-YafQ mutant complex could be co-purified by gel filtration chromatography, demonstrating that the interaction between ␤2 and ␣4 of DinJ (without the contribution of ␣3) and YafQ is sufficient for maintaining a complex assembly. Strikingly, in vitro, the mutant DinJ(I59N/I62N)-YafQ complex cleaved RNA similarly to isolated YafQ. Mutation of residue Phe-49 to Ala in DinJ did not release YafQ from its complex with DinJ, and the RNase function of YafQ was not activated (Fig. 3B). Notably, both mutant complexes were still able to bind a DNA target (Fig. 3D).

Both the DinJ-YafQ Complex and DinJ Alone Only Bind One of the Three Imperfect Palindromic Sequences in the dinJ-yafQ
Operator Region-In the dinJ-yafQ system, three imperfect palindromic sequences (palindromes pal I, pal I-II, and pal II) have been reported previously as putative dinJ-yafQ operator regions (22). Palindromes pal I and II are juxtaposed with one base pair overlapping, whereas pal I-II comprises the 3Ј region of pal I and the 5Ј region part of pal II (Fig. 4A). To validate which palindrome with inverted repeats is the operator region responsible for autorepression regulated by the DinJ-YafQ complex, these three sequences were tested with purified DinJ-YafQ in an EMSA. An intergenic fragment of dinJ-yafQ was included as a control (22). Within the protein concentration range 1-10 M, pal I (CGCTGTTGCTCATTTGAGCTA-CAATT) (reverted repeat underlined) and pal I-II (TTT-GAGCTACAATTCAAGCTGAATAA) bound to the DinJ-YafQ complex, generating distinct upshifts in the observed bands (Fig. 4B, lane 5 and 6), with a migration pattern analogous to the dinJ-yafQ intergenic fragment control. Pal I-II (TTTGAGCTACAATTCAAGCTGAATAA) bound to DinJ-YafQ more specifically than pal I (Fig. 4, C and D). However, pal II, harboring a putative LexA box, did not bind to the DinJ-YafQ complex (only a smear was evident on the gel) (Fig. 4B, lane  7). These results were consistent with an earlier report that demonstrated pal I and pal I-II to be DinJ-YafQ binding fragments (22). To precisely locate the boundaries of the binding sites and define the consensus sequences in pal I and pal I-II for DinJ-YafQ binding, a more comprehensive gel shift analysis was undertaken using gradually shortened DNA duplexes. We synthesized a series of deletion duplexes of pal I-II, and investigated their binding ability to DinJ-YafQ by EMSA. The shortest DNA duplex, which was 11 bp in length (I-II-12, TTTGAGCTACA), was located at the left palindromic sequence in pal I-II and was able to bind to the DinJ-YafQ complex (Fig. 5, A and B). Surprisingly, we were unable to obtain any fragments shortened from the 5Ј-end of pal I-II that were able to bind DinJ-YafQ. Even one duplex, which was only shortened by one base pair at the 5Ј-end of pal I-II (I-II-4), showed a significant loss in affinity to the DinJ-YafQ complex (compared with I-II-7), suggesting the importance of this 5Ј extremity for binding. However, we found that the T-A base pair at the 5Ј-end of pal I-II-7 can be substituted to a C-G base pair (IId7G) with only a slight reduction in binding (Fig. 5C). Other base pairs in this 5Ј region did not contribute significantly to binding affinity. Taken together, our findings indicate that the specific binding site for DinJ-YafQ is located in the 11 bp at the 5Ј-end of pal I-II, with a consensus sequence of TTTGAGCTACA. Because pal I-II partially overlaps with pal I and they share a common sequence of pal I-II-10 (TTTGAGCTACAATT) (Fig.  4A), we next reinvestigated the binding sites in the context of pal I. We questioned whether pal I might provide an additional binding site for DinJ-YafQ, because we occasionally observed two shifted bands at certain concentrations of DinJ-YafQ in EMSAs with pal I (Fig. 4C). Indeed, EMSA with a series of deletion duplexes of pal I showed that one additional deletion duplex, I-3, retained binding ability to DinJ-YafQ, generating a single shifted band (Fig. 5D). This sequence of duplex, I-3 (CGCTGTTGCTCATT), located at the left half-site side of pal I, bears an imperfect reverted sequence of pal I-II-10 (TTT-GAGCTACAATT). Therefore, pal I harbors two half-sites in the dinJ-yafQ operator for DinJ-YafQ regulation, and each of the individual pal I half-sites (I-3 and I-II-10) was able to bind to DinJ-YafQ alone. This explains the observation that pal I-II was also capable of interacting with the DinJ-YafQ complex because pal I-II contains one half-site overlapped by pal I.
The crystal structure of DinJ revealed that it contained an RHH motif in dimeric form in its N-terminal region. This RHH fold is thought to be responsible for binding to each half-site in a palindrome. Next, we addressed whether DNA binding sites are solely recognized by DinJ. We performed similar experiments with all three palindromic candidates. Surprisingly, DinJ was able to bind to pal I and the whole intergenic fragment, but it was unable to bind to both pal I-II and pal II (Fig. 4B). We deduced that pal I contains two half-sites for DinJ-YafQ binding and that DinJ probably requires the presence of two halfsites in pal I simultaneously for binding. The affinity of a halfsite present in pal I-II with DinJ dimer was relatively low; however, a pal I fragment with two half-sites might have a higher affinity for binding DinJ when a dimer-of-dimer DinJ forms. This would be consistent with the observation that DinJ formed a homotetramer in solution. As noted, in contrast to the DinJ-YafQ complex which bound each half-site in the operator separately, DinJ could not bind to a DNA duplex harboring only one palindromic half-site (pal I-II-10 or I-3) in our hands (Fig. 4, A and D) (data not shown). It is probable that YafQ increases the binding affinity of DinJ (in the DinJ-YafQ complex) to individual half-sites in pal I. Overall, we deduced that pal I (positioned at Ϫ62 to Ϫ37 from the translation start site) is the consensus sequence of the dinJ-yafQ operator.
To analyze DNA binding ability in the DinJ mutants, EMSAs were performed with the pal I-II fragment. The results showed that the binding abilities of mutant complexes DinJ(R8A/ K16A)-YafQ, DinJ(S31A/R35A)-YafQ, and DinJ(R8A/K16A/ S31A/R35A)-YafQ were all abolished (Fig. 6C). This indicated that residues Arg-8, Lys-16, Ser-31, and Arg-35 of DinJ are critical for the interaction between the DinJ-YafQ complex and its repressor DNA and that the convex surface in DinJ-YafQ, formed by RHH motifs in the DinJ dimer, constitutes the interacting interface with the operator.

DISCUSSION
In this report, we determined the crystal structure of YafQ in complex with DinJ and the crystal structure of free YafQ. Both structures revealed that YafQ resembles a microbial RNase fold. According to structure-based alignment, all critical residues constituting the enzymatic site of YafQ were highly conserved with those of RNase T1. A recent mutagenesis study on the ribonuclease activity of YafQ demonstrated that His-50, His-63, Asp-67, Arg-83, His-87, and Phe-91 substitutions of the predicted active site residues of YafQ abolished mRNA cleavage in vivo, whereas Asp-61 and Phe-91 mutations inhibited YafQ ribonuclease activity only moderately. This is in agreement with the results revealed by the crystal structure of YafQ. As an endoribonuclease, YafQ enzymatic activity is blocked upon the formation a DinJ-YafQ complex. The coiled linker and helix ␣3 of DinJ in the DinJ-YafQ complex might play critical roles in blockage of RNA access to the catalytic site. Our site-directed mutagenesis studies showed that mutations in helix ␣3 of DinJ conferred ribonuclease activity of the DinJ-YafQ complex. This is in agreement with our expectations that the helix ␣3 mutant of DinJ designed to disrupt the hydrophobic interactions with YafQ would lead to constriction of the Trp-56 loop in YafQ back to the ␤-sheet core of YafQ. Intriguingly, site mutations of the coiled linker region or helix ␣3 in DinJ did not change the formation of the DinJ-YafQ complex, indicating that the primary interactions between DinJ and YafQ were mainly attributed to the C-terminal portion of DinJ. This interaction was shown to prevent YafQ binding to the ribosome and thereby inhibit its ribosome-dependent ribonuclease activity and toxicity in vivo.
Regarding the three imperfect inverted repeat sequences (pal I, pal I-II, and pal II), which overlap in the dinJ-yafQ system, we verified that pal I is the operator for regulating the DinJ-YafQ TA system. DinJ and YafQ assemble into a heterotetramer (YafQ-(DinJ) 2 -YafQ) capable of binding to duplex pal I. DinJ alone was also able to bind to pal I, but it required the presence of two imperfect inverted repeats in the pal I sequence for effective binding. These findings raise the question of what role pal I-II and pal II are playing in transcriptional regulation of dinJ-yafQ and whether they are specific cis-elements that are recognized by other regulatory factors. For example, pal II might be LexA-regulated in the SOS response. Pal I-II might also be recognized by other unidentified regulatory factors for modulating dinJ-yafQ expression, and there may be other TA factors involved in cross-regulation (51). It would therefore be interesting to investigate the possibility of a coordination network for global TA system regulation. Future work may enhance our understanding in this area.