The crystal structures of the DesT–UFA-CoA–DNA and DesT–SFA-CoA complexes were determined, and the data collection and refinement statistics are provided in Table 1. Detailed structural descriptions are provided in Supplementary Results and representative electron density maps are shown in Supplementary Fig. 1. Briefly, the fold of each protomer and the dimeric arrangement are similar to those of TetR8,9, QacR10 and EthR11,12, three members of the TetR repressor family. DesT forms a parallel dimer with each protomer composed of nine α-helices folded into two discrete domains (Fig. 1a). The three N-terminal α-helices create the DNA-binding domain, and the six C-terminal helices create the ligand-binding domain. In the latter domain, helices α4, α5, α7, α8 and α9 are arranged in an antiparallel bundle with helix α6 diagonally crossing the ‘top’ of the bundle forming a platform for the DNA-binding domain. Within the dimer, hydrophobic interactions between the paired helices α8 and α9 that cross at the interface of the ligand-binding domains create a central 4-helical bundle. TetR-like proteins are usually described as existing in two conformations: the DNA-bound ‘uninduced’ form and the ligand-bound ‘induced’ form that does not bind DNA13. This description does not apply to DesT because its function is not determined by the simple presence or absence of ligand. DesT has a defined structural pathway to relay ligand shape to the DNA-binding domains some 40 Å away, and is therefore better described in allosteric terms. The DesT–oleoyl-CoA–DNA ternary complex corresponds to the relaxed (R) state and the DesT–palmitoyl-CoA binary complex represents the tense (T) state.

DesT–oleoyl-CoA is bound to a 30 base duplex DNA containing the 18 base-pair pseudo-palindromic desCB promoter sequence7 (AGTgAACgcttGTTgACT) (Fig. 1a). The DNA-binding N-terminal domain contains a helix-turn-helix motif composed of α2, loop L2-3 and α3 (Fig. 1b). Helix α2 acts as the ‘platform’ helix spanning the cognate major groove and α3 acts as the ‘recognition’ helix penetrating the major groove. The N-terminus of α1 binds across the adjacent minor groove. It has been noted that the N-terminus of α4 within the ligand-binding domain contributes to the binding of DNA in the TetR-repressor family, with the helix dipole and a conserved lysine residue (Lys48 in TetR) engaging the DNA backbone14. However, in DesT, the N-terminus of α4 is 7 Å from the DNA and lacks this conserved lysine. The platform helix α2 specifically engages chain 2, and the recognition helix α3 makes extensive interactions with both DNA chains (Supplementary Fig. 2a). The DesT–DNA interface mainly involves the phosphates of the DNA backbone, except for specific interactions with a pair of guanine bases within each half site. Specifically, the guanidinium groups of Arg36 and Arg51 interact with +3 and −8 guanine bases (with respect to the dyad axis of the cognate sequence) within chains 2 and 1, respectively (Supplementary Fig. 2a). In the crystal structure, the DNA binds 50:50 in both orientations reflecting DesT recognition of the sugar-phosphate backbone and the two palindromic guanines. A crystal structure (not shown) using a fully palindromic sequence confirms this mode of recognition (DNAPP in Table 1).

The DNA is bent by 3.3° toward the protein (Fig. 1a) and the B-DNA helix is slightly deformed at the recognition site (Supplementary Fig. 2b, Supplementary Table 1); the major groove is widened from 11.7 Å to 12.1–13.4 Å and the helical repeat increases from 34 Å to 36 Å, slightly unwinding the DNA.

The L-shaped oleoyl (18:1) chain of the ligand is buried within the hydrophobic interior of the DesT ligand binding domain to become an integral component of the hydrophobic core (Fig. 1c, Supplementary Fig. 3a). The CoA moiety was not visible in the electron density map and is presumably disordered (Supplementary Fig. 1a). The kink in the 18:1 chain at the cis-9 double bond directs the chain towards the periphery where the distal end projects between α4 and α7, and phenylalanines 71, 96, 107 and 166 together with Leu169 form a phenylalanine-rich cluster directly above the kink and below α6. The paired C-termini of α6 and α6′ loosely associate via stacking interactions between the side chains of Arg112 and Arg122′, and an edge-to-face interaction between Tyr115 and Pro170′ at the C-terminus of α8 (Supplementary Fig. 3a).

In the DesT–16:0-CoA binary complex, the entire 16:0-CoA molecule is visible. The C-terminus of DesT is well ordered and reveals the adenine ring interactions. The adenine ring stacks between Ile199 and His205′ adjacent to Trp204′, and the N1 and N6 nitrogen atoms form hydrogen bonds with the amide nitrogen of His205′ and the carbonyl oxygen of His203′, respectively. The entrance to the ligand binding pocket is largely unaffected by the ligand swap because the first 7 carbons of the 16:0 chain form similar interactions to those of the oleoyl chain. However, the tail of the 16:0 linear chain inserts directly into the hydrophobic core and creates a new 9 Å hydrophobic pocket, reminiscent of the ligand-binding hydrophobic pocket in EthR11,12. Most notably, the acyl chain penetrates the phenyalanine-rich cluster below α6 and these residues adjust their positions to accommodate the ligand (Fig. 1d; Supplementary Fig. 3b). These rearrangements result in three conformational changes. First, helices α4 and α7 pack closer together, and a slight bend is induced in α4 due to the local unwinding of residues 68 to 71. This unwinding is associated with the movement of Phe71 and is stabilized by a hydrogen bond between the side chain of Thr70 and the carbonyl oxygen of Glu66. Second, the C-terminus of α6, loop L6-7 and the N-terminus of α7 engage loop L8-9 which extends across the interface and becomes ordered. L8-9 contains five proline residues and appears to act as a clamp that specifically stabilizes the T state conformation. Finally, and most significantly, the C-termini of the paired α6 helices slide closer together by one turn (~4.4 Å), and this new location is stabilized by new interactions centered on Tyr115 (Supplementary Fig. 3b). The side chain engages a hydrophobic pocket comprising Leu108′, Ala111′, Arg112′, Tyr115′, Leu169′, Pro170′ and Ile173′; the main chain carbonyl oxygen becomes hydrogen bonded to the side chain of Arg112′; and the guanidinium group of Arg129 swaps hydrogen-bonding partners from the OH of Tyr115 to the side chain of Gln114.

The core four-helix bundle of the dimer in most TetR family members is a stable substructure14, and this is also true in DesT. Superposition of the paired α8/α9 and α8′/α9′ helices in the R- and T-states clearly shows the three conformational changes described above (Fig. 2a). It also reveals the substantial relative movements of the DNA-binding domains in the T form away from their optimal DNA-binding orientations in the R form (Fig. 2b). The recognition helices α3 and α3′ rotate by some 5°, and their center-to-center distance increases from the required DNA binding distance of 36.7 Å to 41.9 Å. These movements are comparable to those observed in TetR where the distance between the recognition helices increases from 36.6 Å to 39.6 Å8.

The structures suggest specific roles for key residues in transmitting the ligand shape information to the DNA-binding domain (Fig. 2c). Tyr115 at the C-terminus of α6 is important in the T state to fix the orientation of this helix at the dimer interface. We confirmed this by showing that DesT Y115A, unlike DesT (Table 2; Fig. 2d), constitutively binds DNA regardless of the structure of the bound acyl-CoA (Table 2; Fig. 2e). DesT Y115A binds both 16:0-CoA and 16:1Δ9-CoA with similar affinities (Table 2), illustrating that this substantial modification of its DNA binding properties did not simply arise from an inability to bind SFA-CoA. The side chain of Phe71 contacts the ligand to introduce a bend into α4 in the T state, and DesT F71A also exhibits constitutive DNA binding (Table 2). Phe96 and Leu169 are two other side chains that stabilize the hydrophobic core of the T state, and DesT F96A and DesT L169A bind UFA- and SFA-CoAs equally well and constitutively bind DNA (Table 2). We note that F71A, F96A and L169A will each leave voids in the hydrophobic core that could be accessed by the flexible 16:0 acyl chain and contribute to the failure of 16:0 to stabilize the T state.

Phe166 is exquisitely positioned to detect the conformation at the 9-position of the acyl chain where the cis double bond is located (Fig. 2c). This role directly reflects the physiological function of DesT, which is to control the expression of a desaturase that introduces a cis double bond at the 9-position of the acyl chain7. DesT F166A binds both acyl-CoA ligands, but fails to bind DNA under any condition (Table 2; Fig. 2f), and is therefore locked into the T state regardless of fatty acid structure. Phe166 sits directly below α6, and stabilizes its orientation in the R state (Fig. 1c, Supplementary Fig. 3a). The lack of the aromatic side chain prevents the transmission of ligand shape information to α6 leaving the F166A mutant locked in the T state.

These in vitro results were corroborated in vivo by analyzing desCB gene expression in strain PA0482 (ΔdesT) expressing either DesT, or the Y115A or F166A mutants from a plasmid (Fig. 2g). The expression of the wild-type DesT illustrates the normal activity of the protein. In the absence of an exogenous ligand, DesT has repressor activity in vivo due to endogenous ligands that establish an equilibrium between the T and R states. The addition of 16:1Δ9 to the medium represses desCB transcription by stabilizing the R state, whereas the addition of 16:0 to the medium activates transcription through the stabilization of the T state (Fig. 2g). DesT Y115A is permanently locked in the R state in vitro (Fig. 2e), and accordingly, desCB transcription is repressed regardless of the fatty acid presented to the cells (Fig. 2g). On the other hand, DesT F166A is locked in the T state (Fig. 2f), and there is no repression of desCB transcription regardless of the fatty acid in the medium. Cells expressing DesT F166A have desCB mRNA levels that are the same as in cells harboring an empty vector indicating that the F166A mutant cannot repress desCB in vivo (Fig. 2g).

Recombinant DesT proteins with a C-terminal His-tag were purified7. The DesT mutants were generated by PCR based mutagenesis by introducing the base changes for the amino acids using the Quikchange mutagenesis kit (Stratagene Inc.). The identity of the constructs was confirmed by DNA sequencing. The stability of DesT and its mutant derivatives were compared using the SYPRO Orange dye-binding assay19. The transition temperatures for protein unfolding were: DesT, 63°C; Y115A, 55°C; F166A, 61°C; F96A, 52°C; F71A, 56°C; and L169A, 56°C.

An NcoI site was engineered into the multiple cloning site of pUC2020 and the 70-bp NcoI-BamHI fragment from pET15b was introduced to make the Pseudomonas shuttle vector pUCP20-Hy. DesT and the DesT mutants (Y115A and F166A) were cloned into pPUCP20-Hy by digesting the pET28b expression plasmids with BlpI-Klenow filled for blunt ends and NcoI to generate a 735-bp fragment that was inserted into pUCP20-Hy. The resulting plasmids were transformed into P. aeruginosa strain PAO482 (ΔdesT)21, and the strains were grown in M9 + 0.4% glycerol minimal medium in the presence or absence of 0.1% palmitate (16:0) or pamitoleate (16:1Δ9). RNA was isolated using Ambion RNAqueous purification kit (Ambion). The expression levels of desB were measured by real-time PCR5. Values were compared using the C_T method, where the amount of desB cDNA (2^−ΔΔCT) was normalized to rpoD (ΔC_T) and comparisons were made using ΔΔC_T method.

Direct binding of acyl-CoAs to DesT was measured by fluorescence spectroscopy on a Fluorolog-3 spectrofluorimeter (Horiba Jobin Yvon)7. Intrinsic protein fluorescence was measured with excitation at 280 nm and emission at 340 nm (slits set at 5 nm). The concentration of DesT was 4 μM in 20 mM Tris-HCl buffer, pH 7.5, and acyl-CoAs were titrated in 2-μl increments from stock solutions. The data were not corrected for the inner filter effect due to the low absorbance (0.008 average) of acyl-CoA in the experiments22. Each curve was corrected for the nominal fluorescence of acyl-CoA and fitted to one site specific binding equation Y=Bmax×X/(Kd + X), where Y is the fluorescent signal of the protein and X is the acyl-CoA concentration. Examples of the fluorescence titration experiments are provided in Supplementary Fig. 4.

Protein-DNA gel retardation assays were performed using the 32P-labeled desCB oligo probe as described6,7. Assays contained, 20 mM Tris pH 7.5, 50 mM NaCl, 1 mM DTT, 1 mM EDTA, 300 μg/ml BSA, 0.05 unit poly[d(I-C)], and [32P]DNA (2500 cpm, approximately 1 × 10⁻¹¹ M) in 20 μl in the presence of 0, 1 or 10 μM 16:0-CoA or 16:1Δ9-CoA. The mixture was incubated at 22°C for 30 min, and loaded onto a 6% DNA retardation gel. The gels were autoradiographed against a phosphor storage screen and quantified with a Typhoon 9200 (GE Healthcare) using ImageQuant 5.2 software (Molecular Dynamics). DesT binding was indicated by the conversion of the free DNA probe to the DesT–DNA complex. The apparent Kds were determined as described above. Examples of the gel shift experiments are shown in Supplementary Fig. 4.

P. aeruginosa DesT selenomethionine protein was crystallized in the presence of an oligonucleotide corresponding to the desCB promoter (30mer duplex with a 5′ T overhang, 5′-TTACATCAGTGAACGCTTGTTGACTCGATTG) and 18:1Δ9-CoA at 18°C by sitting drop vapor diffusion under mineral oil. The 4.5 μl drop contained 2 μl DesT-ligand mixture (0.3 mM protein, 0.3 mM oligo, 0.3 mM 18:1Δ9-CoA), 2 μl mother liquor (ML) (0.1 M MES pH 7.0, 9% PEG 20K), and 0.5 μl 1 M ammonium sulfate. Crystals were cryo-preserved in 30% glycerol/70% ML. SAD data at the Se peak (0.9792 Å) were collected at the SER-CAT ID beamline to 3 Å. The substructure was determined by HySS23. An initial model of the protein component was generated using Resolve24, and this served as the Phaser25 molecular replacement model for the ternary complex 2.65 Å data collected at 1.0 Å on the SER-CAT BM beamline. The crystals belong to space group I4₁ with one protomer and one DNA strand in the asymmetric unit. The DesT dimer/DNA duplex assembly is generated by 2-fold symmetry. The final model lacks residues 1–3, 175–182 (the L8-9 loop), C-terminal residues 206–209, and tag residues 210–226, and only the buried acyl chain of 18:1Δ9-CoA and DNA bases 5–27 were visible. Ramachandran statistics show that 94.7% and 5.3% of the residues are in the preferred and allowed regions, respectively.

DesT was crystallized in the presence of 16:0-CoA at 18°C by the sitting drop method. The ML contained 0.1 M HEPES pH 7.5, 0.1 M magnesium acetate, and 15% PEG 4K. The drop contained 2 μl DesT-ligand mixture (20 mg/ml protein, 0.9 mM 16:0-CoA) and 2 μl ML. Crystals were cryo-preserved in 15% glycerol/85% ML. A 2.3 Å dataset was collected at 1.0 Å on the SER-CAT BM beamline, and the DesT–18:1Δ9-CoA–DNA model was used for molecular replacement with Phaser. The crystals belong to space group P2₁ with the dimer in the asymmetric unit. The final model lacks residue 1, C-terminal residue 209, tag residues 210–226, residues in the L4-5 loop; residues 83–84 in chain A and 83–85 in chain B. Ramachandran statistics show that 98.5% and 1.5% of the residues are in the preferred and allowed regions, respectively.

Model building was performed using COOT26. Structure refinement was performed using Refmac27 and CNS28, and 5% of the data was sequestered for the calculation of Rfree. Structural figures were generated with PyMol29, and ligand cavity volumes were calculated with Castp30 and HOLLOW31. The final structure statistics were calculated using PROCHECK32. The analysis of the DNA conformation shown in Supplementary Fig. 2b and Supplementary Table 1 was performed using 3DNA33.