Structural and Functional Characterization of a Ketosteroid Transcriptional Regulator of Mycobacterium tuberculosis*

Background: KstR2 regulates cholesterol catabolic genes in Mycobacterium tuberculosis. Results: Dimeric KstR2Mtb binds two molecules of HIP-CoA with high affinity. Each binding site spans the subunits and includes residues conserved in TetR family repressors (TFRs) that bind CoA thioesters. Conclusion: HIP-CoA binding to KstR2Mtb induces a conformation that abrogates DNA binding. Significance: The study identifies molecular determinants of cholesterol catabolism and CoA binding in TFRs. Catabolism of host cholesterol is critical to the virulence of Mycobacterium tuberculosis and is a potential target for novel therapeutics. KstR2, a TetR family repressor (TFR), regulates the expression of 15 genes encoding enzymes that catabolize the last half of the cholesterol molecule, represented by 3aα-H-4α(3′-propanoate)-7aβ-methylhexahydro-1,5-indane-dione (HIP). Binding of KstR2 to its operator sequences is relieved upon binding of HIP-CoA. A 1.6-Å resolution crystal structure of the KstR2Mtb·HIP-CoA complex reveals that the KstR2Mtb dimer accommodates two molecules of HIP-CoA. Each ligand binds in an elongated cleft spanning the dimerization interface such that the HIP and CoA moieties interact with different KstR2Mtb protomers. In isothermal titration calorimetry studies, the dimer bound 2 eq of HIP-CoA with high affinity (Kd = 80 ± 10 nm) but bound neither HIP nor CoASH. Substitution of Arg-162 or Trp-166, residues that interact, respectively, with the diphosphate and HIP moieties of HIP-CoA, dramatically decreased the affinity of KstR2Mtb for HIP-CoA but not for its operator sequence. The variant of R162M that decreased the affinity for HIP-CoA (ΔΔG = 13 kJ mol−1) is consistent with the loss of three hydrogen bonds as indicated in the structural data. A 24-bp operator sequence bound two dimers of KstR2. Structural comparisons with a ligand-free rhodococcal homologue and a DNA-bound homologue suggest that HIP-CoA induces conformational changes of the DNA-binding domains of the dimer that preclude their proper positioning in the major groove of DNA. The results provide insight into KstR2-mediated regulation of expression of steroid catabolic genes and the determinants of ligand binding in TFRs.

. HIP catabolism is largely uncharacterized, but is initiated by FadD3, an acyl-CoA synthetase that transforms the substrate to HIP-CoA (10,11). Our recent results indicate that the KstR2 regulon, which includes fadD3, encodes HIP catabolic enzymes (12).
KstR2 (Rv3557c) is one of two TetR family repressors (TFRs) involved in regulating the cholesterol catabolic genes in mycolic acid-producing Actinobacteria (13). The other one, KstR (Rv3574), is involved in regulating the transcription of genes involved in cholesterol uptake and degradation of the side chain and rings A/B (13,14). KstR2 regulates the expression of 15 genes (Fig. 1B), including fadD3 and ipdAB. IpdAB is required for virulence in Rhodococcus equi and an ipdAB deletion strain has been patented as a live vaccine (13,15). KstR2 binds to three ϳ14-bp inverted palindromic operator sequences, or KstR2 boxes, located at the intergenic regions of the regulon (13). Casabon et al. (12) recently demonstrated that binding of KstR2 Mtb to its operator sequences is relieved by HIP-CoA, the product of the FadD3-catalyzed reaction (10). In contrast, HIP, CoASH, and a variety of cholesterol metabolites did not relieve KstR2 binding to DNA.
TFRs are named after a founding member involved in regulating tetracycline resistance (16) and is one of the most widely distributed families of transcriptional regulators in bacteria (17)(18)(19). Despite significant sequence variation, the proteins form a conserved L-shaped ␣-helical structure featuring an N-terminal DNA-binding domain (DBD) and a larger C-terminal effector-binding domain (EBD) (17). The DBD contains a helix-turn-helix motif involved in binding to operator DNA. TetR protomers associate into homodimers or higher oligomers that recognize palindromic sequences in the operator DNA. More specifically, an ␣-helix of the TetR DBD called the recognition helix forms specific electrostatic and aromatic contacts in the major groove of the DNA of the operator sequence (17). The vast majority of TFRs are repressors in the absence of their effector, whereas the binding of the effector triggers a conformational change that shifts the position of the recognition helix, resulting in release of operator DNA by the regulator.
Despite a generally conserved structure and mechanism of action, the specific position of the ligand binding pocket in TFRs and its chemical composition vary dramatically, resulting in specific responses to a vast assortment of small molecules. This variation also makes it difficult to predict the chemical nature of the cognate ligand, necessitating the characterization of individual family members.
Herein, we used a combination of isothermal titration calorimetry (ITC), electrophoretic mobility assays (EMSA), x-ray crystallography, and directed mutagenesis to characterize the molecular function of KstR2 from M. tuberculosis, KstR2 Mtb . The data define interactions between KstR2 and its effector, HIP-CoA, and provide insights into the function of this regulator in the bacterial catabolism of steroids as well as into TFRs in general.

MATERIALS AND METHODS
Chemicals and Reagents-ATP, CoASH, and cholesterol (Ͼ99%) were purchased from Sigma. NdeI and HindIII Fast Digest restriction enzymes were purchased from Thermo Fisher Scientific Inc. T7 DNA ligase and DpnI were purchased from New England Biolabs. Oligonucleotides were ordered from Integrated DNA Technologies. FadD3 and poly-Histagged tobacco etch virus protease (TEV Pro ) were produced as previously described (10,20). Water for buffers was purified using a Barnstead Nanopure Diamond TM system to a resistivity of at least 18 M⍀. Reagents were of HPLC or analytical grade.
DNA Manipulation-Plasmid DNA was manipulated and propagated using standard procedures (21). Oligonucleotidedirected mutagenesis was performed using the QuikChange TM PCR protocol with slight modifications. Briefly, a single 5Ј phosphorylated mutagenic DNA primer was annealed to pETKstR2 carrying a gene encoding poly-His tagged (Ht-)-KstR2 Mtb (12), then amplified using Phusion DNA polymerase. T7 DNA ligase was added to the reaction mixture to form single-stranded mutagenized plasmid DNA. Template DNA was removed using DpnI and the remaining ssDNA was electroporated into Escherichia coli NovaBlue. The R162M and W166L variants were producing using primers with the following respective nucleotide sequences: 5Ј-pGTCTACCGATTCAT-CATGGACACCACCTGGGTG-3Ј and 5Ј-pCATCCGTGA-CACCACCCTCGTGTCGGTGCGCTGG-3Ј. The nucleotide sequences of variant kstR2 were confirmed.
Purification of KstR2-Wild-type and variant KstR2 Mtb were produced using E. coli Rosetta 2(DE3)pLysS carrying the appropriate derivative of pETKstR2 as previously described (12). The proteins were purified as previously described (12) with the following modification. The affinity purified Ht-KstR2 Mtb was dialyzed overnight against cleavage buffer (25 mM HEPES, pH 7.5, 50 mM KCl, 1 mM DTT, and 0.5 mM EDTA). The affinity tag was removed by incubating ϳ100 mg of Ht-KstR2 Mtb with 0.5 mg of TEV Pro in 10 ml of cleavage buffer overnight at 4°C. Complete digestion was confirmed by SDS-PAGE analysis. TEV Pro -digested KstR2 Mtb was loaded onto Mono-Q 10/100 HR (GE Healthcare) and eluted as previously described (12). Proteins were exchanged into 25 mM HEPES, pH 7.5, 50 mM KCl, concentrated to ϳ20 mg ml Ϫ1 and flash frozen in liquid nitrogen as beads. Typically, 50 mg of protein were purified per 1 liter of culture. Protein concentrations were measured using the bicinchoninic acid (BCA) protein assay with bovine serum albumin as a standard.
Preparation of HIP and HIP-CoA-HIP was obtained using a ⌬fadD3 mutant of RHA1 as previously described (12). HIP-CoA was produced by incubating 2 mM HIP with 2.25 mM ATP, 2.25 mM CoASH, 5 mM MgCl 2 , and 5 M FadD3 in 800 l of 25 mM HEPES, pH 7.5, 50 mM KCl for 30 min. HIP-CoA was purified at room temperature by high-performance liquid chromatography (HPLC) using a Luna 3-m PFP(2) 50 ϫ 4.6-mm column (Phenomenex) in 100 mM ammonium acetate, pH 4.5, at 1 ml min Ϫ1 over a 20-ml linear gradient of 0 -90% methanol. HIP-CoA containing fractions were pooled and methanol was removed under nitrogen. HIP-CoA was purified to Ͼ95% purity and its identity was confirmed by ESI-MS as described (12). HPLC purified HIP-CoA was desalted using a Strata-X 33u 30-mg solid phase extraction (SPE) column (Phenomenex). The solid phase extraction column was equilibrated with 1 ml of methanol, then 1 ml of water. The HIP-CoA solution was passed through the column, washed with 1 ml of water, and eluted in 100% methanol. HIP-CoA was dried under nitrogen and solubilized in 100 l of 25 mM HEPES, pH 7.5, 50 mM KCl. Its concentration was determined spectrophotometrically using ⑀ 260 nm ϭ 11.9 mM Ϫ1 cm Ϫ1 . Typical mole recoveries ranged from 70 to 80%. HIP-CoA for co-crystallization was produced as described above but in a final volume of 6 ml (2.9 mg of HIP). HPLC-eluted fractions containing high concentrations of HIP-CoA were desalted on the solid phase extraction column, dried under nitrogen, suspended in 250 l of water, and lyophilized overnight. The residue was suspended in 50 l of water to a final concentration of 62 mM.
Isothermal Titration Calorimetry-ITC experiments were performed using an ITC200 instrument (GE Healthcare) operated at 25°C and a stirring speed of 1000 rpm. Titrations were performed using 25 mM HEPES, pH 7. titrant. Injections of buffer into KstR2 Mtb and variants showed no significant background heats. The data were processed by subtracting the background heats and removing outlier data points. One-and two-site models were fit using Origin 7.0. Experiments were independently repeated at least three times.
Crystallization of KstR2 Mtb ⅐HIP-CoA-Crystals of the KstR2 Mtb ⅐HIP-CoA complex were obtained by mixing 2 l of 48 mg ml Ϫ1 protein with HIP-CoA at final concentration of 1 mM and 2 l of reservoir solution (0.2 M ammonium sulfate, 0.1 M bis-Tris, pH 5.5, and 25% (w/v) PEG3350) using the hanging drop vapor diffusion method. Crystals appeared at room temperature and were flash frozen in liquid nitrogen after being cryoprotected with paratone oil. X-ray diffraction data were collected at 100 K using a Rigaku HomeLab system featuring Micromax-007 HF rotating copper anode fitted with a Rigaku R-AXIS IVϩϩ image plate detector. Diffraction data were processed and reduced using the HKL-3000 software package (22). The crystal structure was solved by molecular replacement using MrBump from the CCP4 software package (23) with the structure of KstR2 RHA1 (Ro04598) from R. jostii RHA1 (PDB code 2IBD) as a search query. Structure refinement was carried out using Phenix.refine (24) and Coot (25). Geometry was verified using Phenix.refine, Coot, and the RCSB PDB Validation server. The final asymmetric unit contains one copy of the KstR2 Mtb protein chain encompassing residues . The presence of one copy of HIP-CoA in the asymmetric unit was verified using simulated annealing (Cartesian) omit maps using Phenix.refine with default parameters, followed by model building into residual positive F o Ϫ F c density and occupancy refinement of HIP-CoA.
Structural Analysis-The PDBePISA server was used to analyze inter-protomeric and protein-ligand interactions (26). The DaliLite and PDBeFold servers were utilized for structure comparisons (27,28). Electrostatic surfaces were analyzed using Chimera (29). Binding cavity properties were analyzed using the CASTp server (30).
Electrophoretic Mobility Shift Assays (EMSA)-A dsDNA probe of the KstR2 Mtb operator sequence located in the intergenic region of rv3557c and rv3558 was prepared by heating complementary ssDNA oligomers to 95°C and annealing at room temperature in 20 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , and 75 mM NaCl as previously described (12). DNA probes were labeled with DIG-11-ddUTP using the second generation DIG gel shift kit from Roche Diagnostics according to the manufacturer's protocol. Binding assays contained 0 -2 pmol of KstR2 Mtb (WT or variant), 0.04 pmol of DIG-labeled DNA probe, and 0 -10 nmol of HIP-CoA in 20 l of 20 mM HEPES, pH 7.6, 10 mM (NH 4 ) 2 SO 4 , 1 mM DTT, 0.2% (w/v) Tween 20, 30 mM KCl, and 1 mM EDTA. Assays were incubated for 30 min at 37°C, then loaded onto 9% polyacrylamide gels containing 0.5ϫ TBE. Gels were run for 45 min at 105 V, then blotted onto positively charged Hybond-N ϩ nylon membranes (GE Healthcare). DNA was viewed using anti-DIG-alkaline phosphatase and chemiluminescent substrate, CSPD, as described by the manufacturer (Roche Diagnostics). Sequences of DNA probes were 5Ј-GCGTACCAAGCAAGTGCTTGCTTAGGTAGC-3Ј and 5Ј-GCTACCTAAGCAAGCACTTGCTTGGTACGC-3Ј.
Size Exclusion Chromatography-The oligomeric state of KstR2 was analyzed using size exclusion chromatography multiangle light scattering (SEC-MALS). Twenty-five l of 80 M KstR2 Mtb was injected onto a HPLC 1260 Infinity LC (Agilent Technologies) coupled to a Superdex 200 5/150 column (GE Healthcare). A second sample was incubated at 37°C for 30 min with 20 M of a 24-bp DNA fragment representing the KstR2 operator from the intergenic region of rv3549c/rv3550 (13). SEC-MALS was operated at 0.2 ml/min in 25 mM HEPES, 50 mM KCl, pH 7.5. Data were collected using a miniDAWN TREOS multiangle static light scattering device and an Optilab T-rEX refractive index detector (Wyatt Technologies). The molecular weight of complexes was determined using the ASTRA6 program (Wyatt Technologies). The nucleotide sequences of the oligonucleotides used to generate the DNA fragment were 5Ј-ACCTAAGCAAGCACTTGCTTGGTA-3Ј and its complement. Oligonucleotides were HPLC purified by the manufacturer (Integrated DNA Technologies) and annealed as described above in 25 mM HEPES, 50 mM KCl, pH 7.5.

RESULTS
KstR2 Mtb Binds HIP-CoA with High Affinity-We have previously shown that HIP-CoA is the chemical effector of KstR2 Mtb , relieving the binding of the repressor to its operator DNA upon interaction with this molecule (12). We performed ITC to better analyze the interaction of KstR2 Mtb with its effector. The binding was exothermic and driven by enthalpy with an unfavorable entropic contribution (Fig. 2, Table 2). The onesite equation best fit the binding isotherms. No cooperativity was detected and attempts to model the data using a two-site equation yielded poor fits. Replicate titration curves were cen-tered at a mole ratio of 0.89 consistent with a one-to-one stoichiometry between the KstR2 Mtb protomer and HIP-CoA. Under the experimental conditions, the K d was 80 Ϯ 10 nM (25 mM HEPES, pH 7.5, 50 mM KCl).
We then used ITC to test whether KstR2 Mtb binds either HIP or CoASH. Titrations of 400 M CoASH or HIP into 40 M KstR2 Mtb (Fig. 2, C and D, respectively) gave heats that were slightly above background. However, neither compound yielded a titration curve. Increasing the concentrations to 1 mM titrant and 100 M KstR2 Mtb generated a proportional increase in measured heats for both compounds and no titration of CoASH. Titrations at 1 mM HIP were unreliable due to precipitation of KstR2 Mtb during titration.  Table 1.
Similarly to previously characterized TFRs, the KstR2 Mtb protomer adopted an all ␣-helical L-shaped fold covering ϳ25 ϫ 38 ϫ 57 Å (Fig. 3A). The short axis of the protomer comprises the N-terminal DBD (residues 6 to 54) with most ␣-helices in this domain arranged perpendicular to the long axis of the protein representing the C-terminal EBD (residues 55-198) (Fig. 3A). The two domains are connected by a kinked ␣-helix (␣6) with one face of this helix interacting with the DBD and the other with the EBD.
The KstR2 Mtb protomer formed an extended interface of 1687 Å 2 with an adjacent KstR2 Mtb protomer related to the first by a crystallographic 2-fold symmetry axis. This arrangement likely represents the biological dimer, the typical minimal oligomeric state of TFRs. Dimerization of KstR2 Mtb is mediated by contacts between 24 residues, 15 of which are hydrophobic, belonging to ␣-helices ␣8 and ␣9 of the C-terminal domain of each protomer.
The electron density corresponding to the HIP-CoA molecule (Fig. 3B) occupies a large extended cavity (2637 Å 2 in surface area) that spans the two KstR2 Mtb protomers and that sequesters over half of the ligand molecule from the solvent. This cavity, whose shape and chemical nature closely complement that of the ligand, is composed of two elements: a positively charged pocket lined by helices ␣8, ␣9, and their connecting loop in one protomer that binds the adenosine moiety; and a deep hydrophobic pocket defined by helices ␣4Ј, ␣5Ј, ␣6Ј, ␣7Ј, and ␣8Ј in the second protomer that binds the HIP moiety (the prime identifies elements of the second protomer). Thus, each HIP-CoA molecule binds across the KstR2 Mtb dimer interface (Fig. 3), and the two binding clefts are independent of each other. Indeed, the two ligands approach no closer than 7.6 Å.
A total of 23 amino acids from each KstR2 Mtb protomer interact with the HIP-CoA molecule. The adenine moiety of the ligand is anchored primarily through interactions with residues belonging to the loop connecting helices ␣8 and ␣9. The diphosphates of the CoA moiety are stabilized by four hydrogen bonds, including three with Arg-162Ј. The cycloalkanone rings of the HIP moiety form stacking interactions with the aromatic side chains of Trp-166Ј (from ␣8Ј) and Tyr-108Ј (from ␣6Ј) that make up the deep hydrophobic pocket. In addition, the HIP moiety forms many hydrophobic interactions with side chains of residues that line the pocket: Phe-65Ј, Leu-66Ј, Leu-69Ј, Phe-70Ј, Tyr-73Ј, and Val-105Ј. Finally, the 5-carbonyl oxygen forms a hydrogen bond with the side chain of Gln-109Ј. The high number of protein-ligand contacts and the close complementarity between the chemical environment of the binding cleft of KstR2 Mtb and the specific chemical groups of the ligand suggests that KstR2 Mtb is highly specific toward HIP-CoA. This is in line with the biophysical characterization of KstR2 Mtb interactions with this ligand presented above.
Binding of HIP-CoA Alters the Conformation of KstR2-Effector binding typically triggers conformational changes in TFRs. To evaluate whether HIP-CoA binding triggers similar changes in KstR2 Mtb , we sought to structurally characterize the ligand-free form of the regulator. Attempts to obtain crystals of KstR2 Mtb in the absence of HIP-CoA were unsuccessful. However, a structure of the ligand-free form of KstR2 RHA1 from R. jostii RHA1 is available (PDB code 2IBD). KstR2 RHA1 shares 59% amino acid sequence identity with KstR2 Mtb including 19  of the 23 residues that interact with HIP-CoA. A superposition of the KstR2 Mtb ⅐HIP-CoA complex and KstR2 RHA1 structures yielded a root mean square deviation of 1.5 Å over 186 matching C␣ atoms of their protomers (Fig. 4A). Nevertheless, the ligand-binding clefts of these proteins were structurally different (Fig. 4B), particularly surrounding the HIP moiety, suggesting that ligand binding induces the conformational differences observed in the KstR2 Mtb ⅐HIP-CoA complex. The most striking alterations involve the positions of Trp-166 and Tyr-108, which shift up to 4.7 Å to stack on either face of the bicycloalkanone rings of HIP in KstR2 Mtb ⅐HIP-CoA (Trp-170 and Tyr-112, respectively, in KstR2 RHA1 ) (Fig. 4B). The conformational differences include shifts in the positions of helices ␣4, ␣6, and ␣7 that surround the HIP binding pocket and translate to regions beyond. Helix ␣4 connects the N-and C-terminal domains in both KstR2 structures and helix ␣6 forms multiple contacts with ␣1 of N-terminal DBD (Figs. 3 and 4B). Accordingly, the shift in position of these helices leads to a 15°rotation of the DBD in the ligand-bound KstR2 Mtb in comparison to this position of the domain in the ligand-free KstR2 RHA1 structure. In turn, this results in a significant difference in the relative position of the DBDs in the context of the KstR2 dimer (Fig.  4A). These differences are consistent with HIP-CoA binding inducing a conformational change in KstR2.
To evaluate whether the conformational differences between the ligand-bound and ligand-free KstR2 structures would affect the interaction of the regulator with its operator DNA, we com-pared the KstR2 Mtb ⅐HIP-CoA structure with that of a TFR bound to its operator DNA. The closest suitable match retrieved by our PDB search was SlmA from Vibrio cholerae (PDB 4GCT (31)). SlmA and KstR2 Mtb protomers superimposed with a root mean square deviation 2.8 Å over 179 matching C␣ atoms (Fig. 5A) facilitating the identification of potential DNA-binding secondary structure elements in KstR2 Mtb . As is characteristic of TFRs, the SlmA dimer formed symmetric contacts in adjacent major grooves of its operator DNA through helices ␣2 (the recognition helix) and ␣3 in each protomer (Fig.  5). The equivalent region in the KstR2 Mtb structure (between residues 33 and 50) featured prominently exposed residues that could form contacts with DNA (i.e. Val-33, Ser-44 and Tyr-48). Importantly, the relative position of the recognition helices differed dramatically in the operator-bound SlmA and ligandbound KstR2 Mtb dimers (Fig. 5B): in KstR2 Mtb ⅐HIP-CoA, this helix is positioned further away from the DNA major groove. Overall, these analyses suggest that the conformation of the DBD domain in KstR2 Mtb ⅐HIP-CoA is not compatible with binding to its operator. This further suggests that HIP-CoA regulates the DNA-binding activity of KstR2 in the same manner as that established for other TFRs where effector binding induces conformational changes that result in relieving the binding of the TFR to its operator DNA.
Functional Validation of KstR2 Mtb ⅐HIP-CoA Interactions-To functionally validate the KstR2 Mtb ⅐HIP-CoA structural model, two key HIP-CoA binding residues were individually substituted using directed mutagenesis and the resulting KstR2 Mtb variants were characterized using ITC and EMSA. More specifically, the structural data indicate that Arg-162 and Trp-166 form important interactions with HIP-CoA (Fig. 3B) and that their substitution with methionine and leucine, respectively, should disrupt the binding of KstR2 Mtb to its effector but not to its operator DNA.  Isotherms showed that both KstR2 variants were significantly impaired with respect to HIP-CoA binding. The R162M variant bound HIP-CoA with an affinity ϳ200 times lower than WT (Table 2). Like WT, HIP-CoA binding to the R162M variant was enthalpically driven with an unfavorable entropic contribution (Table 2). Unlike WT, the isotherm of the R162M variant showed a shallow titration curve consistent with the variant not being saturated at a 3-fold molar excess of HIP-CoA (Fig. 6A). The one-site equation fit poorly to the isotherm, yielding a stoichiometry of n ϭ 1.77. The W166L variant showed no titration with HIP-CoA: the generated heats were equal to background (Fig. 6B).
Using EMSA, both R162M and W166L variants bound to the KstR2 operator sequence with comparable affinity as WT KstR2 Mtb (Fig. 7). More specifically, WT and variants formed DNA⅐protein complexes at a mole ratio of 1:1 and no proteinfree DNA probe was detected at a ratio of 1:50 DNA:protein. Consistent with previous results, the binding of DNA by WT KstR2 Mtb was relieved in the presence of 50 M HIP-CoA. Consistent with the ITC results, 500 M HIP-CoA was required to detectably relieve binding to DNA by R162M. Moreover, the W166L⅐DNA complex was not detectably disrupted even at high concentrations of HIP-CoA.

A KstR2 Operator Sequence Binds Two KstR2 Mtb Dimers-
The oligomeric state of KstR2 Mtb was investigated using SEC-MALS. KstR2 Mtb eluted as a single peak (t R ϭ 10.2 min; 25 mM HEPES, 50 mM KCl, pH 7.5) with a molecular mass of 42.9 Ϯ 0.2 kDa determined using the Rayleigh ratio (Fig. 8). This is within 10% of the predicted mass of the KstR2 Mtb dimer (46 kDa). The small discrepancy between the two may be partly due to the elongated structure of KstR2 Mtb : the ASTRA6 software calculates molecular weight from molecular radius using spherical structures. To investigate the oligomeric state of KstR2 Mtb bound to its operator, a DNA fragment containing a 14-bp   Ϫ43.0 (0.9) Ϫ66 W166L HIP-CoA NB a No binding detected.

Characterization of KstR2 from M. tuberculosis
KstR2 box flanked by 5 bp on either side was synthesized. A sample of KstR2 Mtb incubated with this 24-bp fragment yielded a single protein-containing peak (t R ϭ 8.5 min) with a molecular mass of 97 Ϯ 1 kDa. This is within 10% of the molecular weight predicted for a complex of two KstR2 Mtb dimers and one DNA fragment (106 kDa). The DNA fragment alone (14.7 kDa) eluted at 10 min and caused negligible light scattering (data not shown).

DISCUSSION
This study provides the first molecular insights into the KstR2-mediated regulation of the expression of steroid catabolic genes in M. tuberculosis and other Actinobacteria. The structural and titration data establish that the KstR2 dimer binds 2 molecules of HIP-CoA. Each HIP-CoA molecule binds in a deep cleft that spans the KstR2 dimer with the adenosine and HIP moieties bound by separate protomers. The extensive electrostatic and hydrophobic interactions that mediate the binding of HIP-CoA to KstR2 Mtb are corroborated by the high affinity of the regulator for its effector molecule. The functional significance of the KstR2 Mtb ⅐HIP-CoA structure was further validated by directed mutagenesis, which established that Arg-162 and Trp-166 contribute significantly to the binding of HIP-CoA, whereas minimally affecting the binding of the operator DNA. Finally, comparison of the KstR2 Mtb ⅐HIP-CoA, KstR2 RHA1 , and SlmA⅐DNA structures suggests how effector binding alters the conformation of the regulator to relieve binding of the operator DNA.
The amino acid sequence conservation among KstR2 orthologs in steroid-degrading Actinobacteria further validates the functional importance of the residues identified in the KstR2 Mtb ⅐HIP-CoA structure. These orthologs share ϳ50% amino acid sequence identity with higher conservation among DBD residues predicted to bind the operator DNA and EBD residues that bind HIP-CoA. Specifically, the residues located on helices ␣2 and 3 in KstR2 Mtb are all conserved with the exception of Gly-39. This is consistent with the conserved nucleotide sequence of the operator across Actinobacteria (13). Similarly, 19 of 23 residues that contact the effector in the KstR2 Mtb ⅐HIP-CoA complex, located between residues 138 and 195, are conserved in KstR2 orthologs.
The occurrence of a single KstR2 regulon in Actinobacteria in strains that contain several distinct steroid catabolic pathways (7,32) suggests that hydroxylated HIP-CoA can act as the effector of at least some KstR2 orthologs. For example, R. jostii RHA1 possesses at least three distinct pathways that converge on the HIP catabolic pathway, two of which are responsible for cholesterol and bile acid catabolism, respectively (32). The catabolism of bile acids such as cholate results in the production of 3Ј-OH HIP (10), suggesting that 3Ј-OH HIP-CoA would be an effector of KstR2 RHA1 . Inspection of the KstR2 Mtb ⅐HIP-CoA structure indicates that there is sufficient space adjacent to C3Ј of the HIP moiety to accommodate a hydroxyl group. Moreover, all of the residues within a radius of 6 Å of C3Ј are conserved in KstR2 RHA1 (Fig. 3B). By contrast, the binding of 7␤-OH HIP-CoA is predicted to be sterically hindered by Tyr-108 (Tyr-112 in KstR2 RHA1 ). This is consistent with the finding that C12 hydroxyl groups of bile acids (corresponding to C7 of HIP) are removed prior to rings C/D degradation in Actinobacteria (10,11). By contrast, 7␤-OH HIP is produced in steroiddegrading Gram-negative bacteria such as Pseudomonas putida DOC21 (33). However, the HIP catabolic genes in these strains appear to be regulated by a LuxR-type transcriptional repressor (33,34).
The affinity of KstR2 Mtb for its effector is in line with what has been reported for other TFRs that bind CoA thioester effectors to regulate catabolic genes. Thus, PaaR and FadR from Thermus thermophilus HB8 are involved in the catabolism of phenylacetate and fatty acids, respectively, and bind phenylacetate-CoA and lauroyl-CoA with K d values of 24 and 90 nM, respectively (35)(36)(37). The regulation of HIP, phenylacetate, and fatty acid catabolism share a common logic because catabolism is initiated by an acyl-CoA synthetase (ligase) whose product is the effector for the corresponding TFR in the pathway. TFRs that bind smaller molecules typically have dissociation constants in the 1-10 M range (18,38). The tighter binding of the CoA thioesters is consistent with the increased number of protein-ligand interactions afforded by the CoA moiety. An interesting exception is DesT, which binds palmitoyl-CoA with a K d value of ϳ3 M (39). However, DesT regulates the unsaturated: saturated ratio of acyl chains in lipid bilayers and binds the acyl-CoA in a fundamentally different way than the other TFRs.
Comparison of the structures of FadRs from T. thermophilus (PDB codes 3ANP and 3ANG (36)) and Bacillus subtilis (PDB code 1VI0 (40)) in complex with fatty acyl-CoA are strikingly similar to the KstR2 Mtb ⅐HIP-CoA complex: in each, the ligand is bound in a cleft that spans the two protomers with the fatty acid and adenosine moieties bound to separate chains. Conserved residues include Arg-159 and Arg-173 (KstR2 numbering) that interact with the adenine and Arg-162Ј, which hydrogen bonds with the diphosphate moiety of CoA. Nevertheless, the adenine ring in the FadR⅐fatty acyl-CoA complexes is flipped 180 o with respect to that in the KstR2 Mtb ⅐HIP-CoA complex and oriented perpendicular to the rotational axis of the dimer. By contrast, the fatty acyl-CoA binds in a different way in the DesT com-plexes, with the CoA moiety at the top of the EBD and the acyl chain extending down a channel, parallel to the ␣-helices of the EBD (39). Consistent with the different binding mode, none of the Arg residues are conserved in DesT. The three conserved Arg residues are also not conserved in TFRs that bind smaller effectors, such as QacR of Staphylococcus aureus (PDB code 1JT6), P. putida TtgR (PDB code 2UXI), and Streptomyces coelicolor ActR (PDB code 3B6A).
Conservation of the CoA-binding residues in KstR2 and FadR may be extended to other TFRs to gain insight into their respective effectors. Comparison of sequence topology maps of 48 structures of TFRs identified by Yu et al. (18) indicate that 13 contain at least two of the three conserved basic residues in orientations permissive to CoA binding. Of note, Fad35R of M. tuberculosis (PDB code 4G12) regulates the expression of fad35, which encodes an acetyl-CoA synthetase involved in fatty acid degradation (41). Although Fad35R binds tetracycline, other evidence suggested that a fatty acyl-CoA could be the physiological effector (41). The presence of Lys-184 on the ␣8-␣9 loop, Arg-166 on the apical end of helix ␣8, and His-170 in the middle of helix ␣8 supports this hypothesis. T. thermophilus HB8 PfmR (PDB code 3VPR), which is predicted to regulate phenylacetate or fatty acid catabolism possess basic residues on the ␣8-␣9 loop and middle of helix ␣8 (Arg-163 and Arg-149, respectively) but lacks a basic residue at the top of helix ␣8. Although the respective effectors of these 13 TFRs have yet to be identified, it appears that a significant subset of TFRs bind CoA thioesters.
Our model predicts that the effector of M. tuberculosis KstR (PDB code 3MNL) is also a CoA thioester. KstR2 Mtb and KstR Mtb share only 18% amino acid sequence identity. However, the conserved residues include four that mediate CoA binding in KstR2 (Arg-162, Asp-163, Trp-171, and Arg-173) and two that mediate steroid binding (Trp-166 and Phe-70). Although the effector of KstR Mtb has yet to be identified, the possibility that it is a CoA thioester is consistent with cholesterol catabolism. More particularly, C27 of the alkyl side chain of the steroid is oxidized to a carboxylate by either Cyp125 or -142 (4) and is then thioesterified by FadD19 (42). It is unclear whether these transformations occur prior to oxidation of the 3␤-hydroxyl. Nevertheless, the role of a cholesten-26-oyl-CoA as the effector of KstR Mtb would mirror the regulatory logic of the KstR2 regulon in M. tuberculosis in that the catabolic genes are induced by the first CoA thioester catabolite produced. Interestingly, 3-oxo-4-cholesten-26-oic acid was identified as the effector of KstR of Mycobacterium smegmatis (43). It seems unlikely that the two KstR orthologs have different effectors considering that they share 87% amino acid sequence identity, including the four residues predicted to interact with the CoA moiety. Indeed, Garcia-Fernandez et al. (43) recognized the possibility that the physiological effector of KstR is a CoA thioester but were unable to test this hypothesis.
The proposed mechanism of response of KstR2 Mtb to HIP-CoA is similar to what has been proposed for other TFRs where the binding of a small molecule effector induces conformational changes that abrogate the interactions of the regulator with its operator DNA (17,18). In the absence of a KstR2 Mtb ⅐DNA structure, the comparative analysis between the KstR2 Mtb ⅐HIP-CoA and SlmA⅐DNA complexes unveiled differences in the relative conformation of DBD domains in the TFRs that, in case of KstR2 Mtb , are likely triggered by HIP-CoA binding. More specifically, HIP-CoA binding to KstR2 Mtb repositions helices ␣4 and ␣6 causing a net 15 o outward rotation of DBD helix ␣1, displacing helices ␣2 and ␣3 compared with corresponding elements in SlmA⅐DNA complex (Fig. 4). Interestingly, Tyr-48 and His-50 on DBD helix ␣3 in KstR2 Mtb ⅐HIP-CoA are rotated outward by ϳ10 Å as compared with the ligand-free structure of KstR2 RHA1 (Fig. 4). These residues are highly conserved in TFRs and, in E. coli TetR, directly interact with the bases in the major groove of the DNA (18). Last, Lys-54, which is also highly conserved in TFRs (18), is displaced by ϳ7 Å in KstR2 Mtb ⅐HIP-CoA compared with KstR2 RHA1 . The effect of HIP-CoA on KstR2 is also remarkably similar to the displacement of helix ␣4 and outward rotation of helix ␣3 observed in apo -versus ligandbound B. subtilis FadR (35). In TetR, effector binding thermodynamically stabilizes the DBD in a conformation that is incompatible with DNA binding, preventing the DBD from assuming a conformation that is competent for DNA binding (44). The inability of KstR2 Mtb ⅐HIP-CoA to bind DNA likely has the same mechanistic origin.
The binding of two KstR2 Mtb dimers to its operator was somewhat unexpected considering that the KstR box of 14 bp (13) is too short to accommodate four recognition helices in the major groove. Typically the TFR operators that bind two dimers are at least 22 bp in length (31,(45)(46)(47). By contrast, TFRs that bind operator sequences of less than 17 bp bind as a single dimer (46). It is possible that the KstR2 box extends beyond the 14 bp identified by Kendall et al. (13). Alternatively, the KstR2 dimers may bind opposite sides of the DNA helix, as in the SlmA⅐DNA complex (PDB code 4CGT). In such a scenario, the first KstR dimer induces a conformational change in the DNA that facilitates the binding of the second dimer to a non-canonical sequence. Intriguingly, KstR2 Mtb contains the Arg-X-Thr motif that is present in SlmA and induces a kink in the DNA (31). Similar models have been invoked to explain cooperative binding in each of two other TFRs, QacR and CprB (46,47), although QacR lacks the Arg residue and does not induce a kink in the DNA. Interestingly, single KstR2 boxes occur between divergently transcribed promoters (13) and gel shift assays indicate that KstR2 binds to each of the three boxes with the same stoichiometry (12). The binding of two KstR2 dimers to opposite sides of the DNA helix may enable the repressor to act at both promoters. Additional experiments and structural data are required to determine the precise architecture of the KstR2 Mtb ⅐operator complexes and how this regulates divergently transcribed promoters.
This first crystal structure of a steroid catabolite-bound TFR provides general insights into the regulation of steroid catabolism in Actinobacteria. Significantly, the catabolism of HIP is poorly elucidated despite the occurrence of the virulence factors such as ipdA in the pathway. Further elucidation of HIP catabolism and its regulation should facilitate the development of novel therapeutics to treat tuberculosis.