Crystal Structure of PhnH: an Essential Component of Carbon-Phosphorus Lyase in Escherichia coli

Cloning and expression of phnH and purification of recombinant PhnH.The gene encoding PhnH was amplified by PCR using a single E. coli K-12 colony (XL1-Blue; Stratagene) as a template. Taq DNA polymerase was used with oligonucleotides 5′-CGT TGG ATC CGA CGA CGA CGA CAA AGC CAT GGG TAT GAC CCT GGA GGC CGC TTT TA-3′ and 5′-TAA TTA AGC TT T TAC TAT CAG CAC ACC TCC ACA TGA GTG G-3′ as forward and reverse primers, respectively, thereby introducing flanking BamHI, NcoI, and HindIII restriction sites (underlined) and three stop codons (bold type). The PCR product was cloned into the NcoI-HindIII sites of the expression plasmid pQI-pD, resulting in pQI-phnH. This construct had an N-terminal hexahistidine tag sequence (MRGSH₆GSGSMG) immediately prior to the first methionine codon of PhnH in GenBank database entry NP418524. pQI-pD is a derivative of pQE-80L (Qiagen) that is under control of a bacteriophage T5 promoter/lac operator element and specifies ampicillin resistance. The correct phnH nucleotide sequence of positive clones was confirmed by sequencing of both DNA strands.

E. coli BL21 cells (Novagen) transformed with plasmid pQI-phnH were grown at 30°C in Lennox broth media (Fisher Biosciences, Canada) supplemented with ampicillin (100 μg ml⁻¹). At an optical density of 0.7 at 600 nm expression of PhnH was induced with 0.5 mM isopropyl-1-thio-β-d-galactopyranoside (IPTG). After incubation for at an additional 4 h at 30°C, the cells were harvested by centrifugation and stored at −20°C. Frozen cell pellets were resuspended in 50 mM imidazole-300 mM NaCl (pH 7) and lysed using an EmulsiFlex-C5 homogenizer (Avestin, Canada). The lysed cells were then centrifuged at 40,000 × g for 30 min at 4°C. The supernatant was collected and applied to a column containing Ni-nitrilotriacetic acid resin (Ni Sepharose high performance; GE Healthcare, United States) preequilibrated with cell lysis buffer. The captured PhnH was then eluted with an imidazole gradient (50 to 500 mM imidazole) over 10 column volumes at a flow rate of 5 ml min⁻¹ using an AKTA fast protein liquid chromatography system. Fractions containing >95% pure PhnH, as shown by sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis, were pooled, concentrated by ultrafiltration (Amicon), and then dialyzed into 50 mM HEPES-300 mM NaCl-1 mM dithiothreitol (pH 7.5). The concentration of PhnH was calculated from the absorbance at 280 nm using the extinction coefficient 11,125 M⁻¹ cm⁻¹ calculated from the amino acid sequence. PhnH was also produced as a selenomethionine derivative in the methionine auxotroph strain DL41(DE3) grown in LE Master medium (11). The selenomethionine-labeled PhnH was purified using the procedure that was used for the native protein. Quantitative incorporation of selenomethionine was confirmed by matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS).

MALDI-MS and size exclusion chromatography of PhnH.The mass spectrum of PhnH was obtained with a Voyager-DE STR MALDI-time of flight mass spectrometer (Applied Biosystems) using sinapinic acid as an ionization matrix. Size exclusion chromatography was performed with an AKTA fast protein liquid chromatography system using a Superdex 200 column (1 by 30 cm; prep-grade; GE Healthcare, United States) equilibrated at a rate of 1 ml min⁻¹ with 50 mM sodium phosphate-150 mM NaCl (pH 7.0). The column was calibrated by injecting 50 μl portions of apoferritin (440 kDa; 2.5 mg ml⁻¹), β-amylase (200 kDa; 4 mg ml⁻¹), alcohol dehydrogenase (150 kDa; 3 mg ml⁻¹), cytochrome c (13.6 kDa; 1.5 mg ml⁻¹), bovine serum albumin (66 kDa; 8 mg ml⁻¹), and carbonic anhydrase (29 kDa; 3 mg ml⁻¹). The void volume was determined by injecting 50 μl of blue dextran (2 mg ml⁻¹). All standards were obtained from Sigma-Aldrich (Oakville, Ontario). PhnH (50 μl of a 5-mg ml⁻¹ solution) was applied to the column and eluted under the same conditions.

Crystallization and data collection.All crystallization reagents were purchased from Hampton Research. Crystals of PhnH were grown for 1 to 2 days at room temperature (∼21°C) using the hanging drop vapor diffusion method. The optimized conditions consisted of mixing 506 μM PhnH (11.5 mg ml⁻¹) in 50 mM HEPES-300 mM NaCl (pH 7.5) at a 1:1 ratio (2 μl plus 2 μl) with crystallization buffer containing 0.2 M magnesium acetate, 0.1 M sodium citrate (pH 2.83), and 11 to 16% polyethylene glycol 4000. The crystallization conditions used for the selenomethionine protein were identical to those used for the native protein. All X-ray data were collected at 100 K. For data collection, crystals were briefly equilibrated in the same crystallization buffer containing 20% glycerol and then flash frozen in liquid nitrogen. Multiple anomalous dispersion data were collected at the X6-A beamline at the National Synchrotron Light Source (Brookhaven National Laboratory, Upton, NY) equipped with an ADSC quantum-210 charge-coupled device detector. Data were processed with DENZO and SCALEPACK (21). All data to 1.57 Å were indexed and scaled at this stage; however, the resolution was truncated for heavy atom substructure solution and structural refinement to ensure reasonable data intensity statistics for the highest-resolution shell.

Structure solution and refinement.The structure of PhnH was determined by the multiple anomalous dispersion method using a selenomethionine derivative at 2.0 Å. The heavy atom positions for three of five selenium atoms were determined and refined using SOLVE (28). Density modification, phase extension to 1.77 Å, and initial model building were carried out in RESOLVE (26, 27). At the peak wavelength, data for the highest-resolution shell (1.86 to 1.77 Å) were 91% complete with an R _merge of 32.3% and an I/σ(I) of 5.6. Manual model building was performed with XFIT/XTALVIEW (5) and subjected to refinement cycles using CNS (4) and REFMAC5, employing default program values with regard to geometric constraints (20). The final model contained one monomer in the asymmetric unit, with 186 water molecules, four sodium ions, and a single acetate ion. No density could be observed for the histidine purification tag, residues 1 to 10, residues 111 to 119, or residues 141 to 145 of the native protein. All crystallographic figures presented here were generated with PyMol (http://www.pymol.org ).

Ligand docking with FlexX.Compound libraries were obtained from the Developmental Therapeutics Program at the National Cancer Institute, National Institutes of Health (http://dtp.nci.nih.gov/ ). Docking of compounds from the diversity and mechanistic compound libraries was performed using the FlexX program of SYBYL (Tripos Inc., St. Louis, MO). All default values were used, and the entire molecule was selected as the target docking surface. Within each compound library, the docking results were scored and sorted automatically using CScore (Tripos Inc., St. Louis, MO).

Ligand and substrate screening.Isothermal titration calorimetry (ITC) was performed using a VP calorimeter (MicroCal Inc., Northhampton, MA). PhnH was dialyzed overnight into 50 mM HEPES-150 mM NaCl (pH 7.5). Ligand samples were prepared in buffer saved from the dialysis of PhnH. All samples were passed through 0.45-μm filters and extensively degassed with stirring prior to use. Titration was performed by injecting 10-μl aliquots of 2.5 mM ligand into the ITC sample cell (1.5 ml) containing PhnH (200 μM) equilibrated at 30°C. The following compounds and ions were tested by ITC: methyl phosphonic acid, ethylphosphonic acid, 2-aminoethylphosphonic acid, phenylphosphonic acid, α-d-ribofuranosyl ethylphosphonic acid, ATP, ADP, AMP, GTP, 3′-5′-cAMP, adenosine, acetyl-CoA, pyridoxal phosphate, pyridoxal, flavin adenine dinucleotide, flavin mononucleotide, NAD⁺, NADP⁺, thiamine diphosphate, S-adenosylmethionine, Ca²⁺, Mg²⁺, Zn²⁺, Ni²⁺, Mn²⁺, Co²⁺, and Fe³⁺. Compounds were purchased from Sigma-Aldrich (Oakville, Ontario); the exception was α-d-ribofuranosyl ethylphosphonic acid, which was synthesized as described previously (16). ³¹P nuclear magnetic resonance (³¹P-NMR) analysis of methylphosphonic acid in the presence of PhnH was performed with a Bruker 300-MHz NMR spectrometer at 25°C and referenced to phosphate. Spectra were acquired for PhnH (50 μM) and methylphosphonic acid (2 mM) in 50 mM HEPES (pH 7.5) following overnight incubation at room temperature.

High-throughput analysis of ligand binding to PhnH was analyzed by following temperature-dependent protein aggregation using static light scattering (StarGazer; Harbinger Biotech, Toronto, Canada) essentially as previously described (23). As ligands, we used 430 common metabolites, amino acids, metals, sugars, nucleotides, and antibiotics (all available from Sigma-Aldrich, Oakville, Ontario). Reactions were performed in 384-well plates (Nunc, VWR, Canada) in 50-μl aliquots containing 0.1 M HEPES buffer (pH 7.5), 0.15 M NaCl, 1 mM ligand, and 15 μg PhnH. To prevent evaporation of water, the samples were covered with 50 μl of mineral oil. The temperature range used was 27 to 80°C.

Microbial methods for analysis of E. coli gene mutants.The E. coli K-12 strains used are shown in Table 1. The growth media used were phosphate-buffered AB minimal medium and Tris-buffered phosphate-free 00P medium with the indicated phosphorus sources. Glucose (0.2%) served as the carbon source, and thiamine was added at a concentration of 1 mg liter⁻¹. Cell growth was measured by determining the optical density at 436 nm with an Eppendorf PCP6121 spectrophotometer (Eppendorf, United States). An optical density at 436 nm of 1 (1-cm light path) corresponded to approximately 3 × 10¹¹ cells liter⁻¹. Utilization of methylphosphonate as the phosphorus source was analyzed by growing cells in 00P medium supplemented with methylphosphonate (2 mM). Parallel cultures with inorganic phosphate (2 mM) or no phosphate served as controls. Bacteriophage P1 transduction was performed as described previously (19) with selection for chloramphenicol resistance at a chloramphenicol concentration of 30 μg liter⁻¹ in NZY medium. To measure the methane production from methylphosphonate, cells were grown in AB minimal medium supplemented with methylphosphonate (2 mM). One-milliliter cultures were incubated in 5-ml serum flasks closed with butyl rubber. After 24 h of incubation at 37°C, flasks were shaken, and a 1-ml sample of the headspace in each flask was removed and subjected to gas chromatography with a Hewlett Packard 5890 gas chromatograph equipped with a Hayesape 80- to 100-mesh column (Grace Davison Discovery Sciences, United States) coupled to a flame ionization detector. Nitrogen was used as the carrier gas at a flow rate of 20 ml min⁻¹.

Protein structure accession number.The structure factors and atomic coordinates determined in this study have been deposited in the Protein Data Bank under accession number 2FSU.

Purification and biophysical characterization of PhnH.PhnH was expressed in soluble form in E. coli with an N-terminal hexahistidine tag and was purified in one step by immobilized metal affinity chromatography (see the supplemental material). PhnH was predicted to be a 22,757-Da protein as cloned and expressed in this study. This was confirmed by MALDI-MS, where a peak at m/z 22,756 was observed in the spectrum, corresponding to the [M-H]⁺ ion of PhnH. In solution PhnH forms a stable homodimer, as calculated from its relative elution volume from a calibrated size exclusion column (see the supplemental material).

As previously described, the phn operon has been extensively studied at the genetic level. However, the low sequence identity shared by several genes in this cluster, including phnH, has precluded identification of putative in vivo functions for individual proteins encoded in this region. As a result, the PhnG and PhnJ proteins are currently classified as hypothetical proteins. Sequence analyses have identified the phn operon in a number of prokaryotic genomes (7, 13), and all examples of this operon have sequences homologous to these hypothetical proteins. Sequence alignments of representative members of the PhnH family, as classified in the Pfam database (http://www.sanger.ac.uk/Software/Pfam/ ), are shown in Fig. 1. It is important to note that while these sequences exhibit a reasonable degree of similarity, there are relatively few strictly conserved residues among putative PhnH homologues. While further discussion of the location of these strictly conserved residues is presented below, it is important to note that several of these strictly conserved residues are localized to a single region of the protein, which we refer to as the putative ligand binding site.

• Open in new tab
• Download powerpoint

FIG. 1.

Sequence alignment of PhnH (accession no. P11686) from E. coli with various homologues from the Pfam database (http://www.sanger.ac.uk/Software/Pfam/ ). The proteins are identified by their Swiss-Prot accession numbers. PhnH homologues from the following organisms were employed in the ClustalW sequence analysis (http://www.ch.embnet.org/software/ClustalW.html ): Shigella dysenteriae serotype 1 (accession no. Q329H9), Rhizobium loti TaxID 381 (Q98GG3), Pseudomonas aeruginosa (Q9HYM2), Agrobacterium tumefaciens strain C58 (Q8UIW1), Mesorhizobium loti strain R7A (Q8KJ86), and Anabaena sp. strain PCC7120 (Q8YUV7). Homologues with lower levels of sequence identity were included to facilitate deduction of the most highly conserved residues in putative PhnH sequences. Strictly conserved residues are indicated by bold type. Secondary structural elements are shown above the corresponding sequence, and the colors match the colors in the structural representation of PhnH shown in Fig. 2. There are vertical lines for every 10 residues of the PhnH sequence.

Crystal structure of PhnH.The three-dimensional crystal structure of PhnH from E. coli K-12 was determined using the multiple anomalous dispersion method and refined to 1.77 Å (Fig. 2). The model was refined to R and R _free values of 19.1 and 24.8%, respectively, with 89.9% of the residues in the most favorable regions of the Ramachandran plot and 10.1% of the residues in allowed regions (Tables 2 and 3). The overall fold is an α/β group fold and is comprised of three layers: the alpha layer, the beta layer, and the mixed alpha/beta layer (Fig. 2).

• Open in new tab
• Download powerpoint

FIG. 2.

Crystal structure of PhnH. The overall fold is a fold of the α/β group and is comprised of three layers. The alpha layer is pink, the central β-sheet is blue, and the mixed α/β layer consists of magenta helices and green β-strands. The right panel provides a rotated perspective 90° relative to the left panel. The N and C termini are indicated. Missing loop regions are indicated by dashed lines in the right panel.

The tertiary structure of PhnH is compact, and all secondary structural components contribute to the formation of a single core domain. The all-alpha layer consists of three α-helices ranging from 10 to 13 residues long. Individual helices (designated helices A, B, and C) in this layer are interspersed with a single β-strand contributing to the central β-sheet that results in an A-β1-B-β2-C-β3 organization in this region of the domain. Also apparent is a single-turn 3₁₀ helix that resides within the coil region connecting strand β2 to the C helix. In addition to the classical hydrogen bonding patterns observed for the individual helical components of this layer, the positions of the helices relative to the core β-sheet are further stabilized by hydrogen bond formation between the residue side chains of helix A and B residues and several side chains of the core β-strand residues. Two foci of hydrogen bonding can be readily identified in the positional stabilization of helices A and B relative to one another and the core β-sheet. In the first instance hydrogen bonding between the side chains and backbone atoms of Ser17, Gln36, Gln50, and Thr54 results in bridging of the N terminus of helix A to strand β1, the adjacent face of helix B, and the coil region connecting β1 to helix B (see the supplemental material). The second network anchors the N-terminal end of helix B to strands β8 and β9 through hydrogen bond interactions between the Ala47, Thr48, Gln126, Thr177, and Gly179 residues distributed in this region. This region is described in further detail below. In contrast, helix C exhibits no hydrogen bonding with the core β-sheet but rather is stabilized as a result of hydrogen bonding in the coil regions bounding its termini. Here, the backbone carbonyl oxygen of Leu77 from helix C forms a hydrogen bond with the amide nitrogen and the OG1 atom of Thr81 found in the coil connecting the helix to strand β3 of the core β-sheet.

The central layer of the domain is composed of a mixed seven-strand β-sheet that is predominantly parallel; the exception is strand β9, which lies antiparallel to the other strands. The core of the molecule is comprised primarily of small nonpolar residues with very few aromatic residues. Therefore, virtually no hydrogen bonding with amino acid side chains residing within the defined regions of the β-strands is observed. The only two exceptions are Gln126 and Thr177, which participate in a large hydrogen-bonding network that is centered about Asn45, located at the N terminus of helix B. The ND2 atom of Asn45 forms a hydrogen bond with OG1 of Thr177. The carbonyl oxygen of Asn45 forms a hydrogen bond with both the amide nitrogen and OG1 atoms of Thr48. The Asn45 amide nitrogen also forms a hydrogen bond with the NE2 atom of Gln126. Additionally, hydrogen bonds are also apparent between the OG1 atom of Thr177 and the OG1 atom of Thr48, between the NE2 atom of Gln126 and the OG1 atom of Thr48, and between the OE1 atom of Gln126 and the carbonyl oxygen of Pro43 (see the supplemental material). This extensive network appears to confer significant rigidity to the coil regions connecting the alpha and beta layers of the PhnH domain.

The third domain layer, containing both alpha and beta secondary structures, is the most unusual feature of the PhnH structure. This layer is comprised of a three-strand antiparallel β-sheet that is formed by strands β6, β7, and β10, as well as two helices, designated D and E. While structural similarity searches have demonstrated some similarity between PhnH and previously determined protein structures for the alpha and beta layers, we have been unable to identify structural matches for models containing the core β-sheet and the mixed α/β third layer. The individual secondary structural elements in this layer are relatively self-contained, exhibiting very few hydrogen bonding interactions between the helices and the three-strand β-sheet. With the exception of the intrinsic hydrogen bonds involved in secondary structure formation, only two hydrogen bond networks between the secondary elements of the alpha/beta layer and regions of the core β-sheet are apparent. In the first region, the main chain carbonyl of Leu109 at the C terminus of helix D forms a hydrogen bond with the amide nitrogen of Ala 121 found at the N terminus of strand β5. Additional stability in this area is conferred by the hydrogen bonding of the OG atom of Ser110 to the main chain oxygen of Leu170, found in the lengthy coil region connecting helix E to strand β8 of the core β-sheet.

The second, and more extensive, network of hydrogen bonds between the β-core and the mixed structural layer is centered about Arg164 (see the supplemental material). This residue is situated at the C terminus of helix E and interacts with main chain atoms in both strand β10 of the alpha/beta layer and strand β9 of the core β-sheet. Here, the NH2 of Arg164 forms hydrogen bonds to both the carbonyl oxygen of Thr189 of β10 and the carbonyl oxygen of Pro186 at the C terminus of strand β9 in the core β-sheet.

The majority of coil regions are also well defined for electron density due to the extensive hydrogen bond networks formed at the termini of the secondary structural elements. One example of this is found at the C terminus of helix B. In this region a type I β-turn is formed, creating a well-ordered connection to strand β2. In addition to the hydrogen bond between the carbonyl oxygen of Asp57 and the hydrogen of Thr60, the turn is stabilized both by main chain-side chain and interside chain interactions (see the supplemental material). A hydrogen bond is formed between the backbone nitrogen atom of Asp57 and the OG1 atom of Thr60. The Asp57 residue also contributes a hydrogen bond between its OD1 atom and the backbone nitrogen atoms of Asn58 and Asp59. Finally, the side chains of Asn58 and Asp59 are stabilized through the hydrogen bonding of the ND2 atom of Asn58 to the OD1 atom of Asp59. Not surprisingly, no density is apparent for the histidine purification tag or the initial 10 amino acid residues of the native protein. Residues 111 to 119 and 141 to 145 were also disordered in the crystal structure.

Crystallographic dimerization of PhnH.As previously described, size exclusion chromatography has demonstrated the dimerization of PhnH in solution. While the asymmetric unit of the PhnH crystal contains a single molecule, an apparent dimer is formed through interactions with a neighboring symmetry-related molecule, defined by the symmetry operators [-x, −z, z+1/2]. The dimerization interface has a buried surface area of 1,173 Ų and is characterized by swapping of the A helices between monomers with the helices oriented in parallel (Fig. 3). The interaction is predominantly hydrophobic and is defined by the face displaying residues Leu22 and Phe18. Further stabilization is contributed by intermolecular hydrogen bonds between the N terminus of the A helix and the turn connecting helix B to strand β2. In this region, both NH1 and NH2 of Arg19 hydrogen bond with the OD2 atom of Asp57 from the other monomer. The NH2 of Arg19 also exhibits intermolecular hydrogen bonding to the carbonyl oxygen of Leu55. Additionally, a hydrogen bond is observed between the side chains of Asp13 of the first molecule and the Asp57 residue of the second monomer. It is important to note that while helix A is topologically considered a member of the all-alpha layer in the PhnH structure, it is angled approximately 60° upward relative to the orientation of helices B and C in this layer. The resulting orientation of the dimer mediated by the interaction of helix A places the three-strand β-sheets of the mixed α/β layer opposite one another.

• Open in new tab
• Download powerpoint

FIG. 3.

Crystallographic dimer is a dimer of PhnH and a symmetry-related molecule. Swapping of the A helices from the two monomers is apparent in the top view shown in the left panel. Rotation of the dimer by 90° highlights the parallel arrangement of the monomers relative to one another, as well as the creation of a pocket in the front face (right panel).

The dimerization of PhnH in this orientation results in the formation a large pocket on the surface of each monomer (Fig. 4). This pocket is bounded by the N termini of strands β4, β5, and β8, the Leu21 face of helix A to the rear, and the Arg19 face of helix A from the other monomer. The pocket is predominantly hydrophobic; two patches of positive charge are contributed by Arg20 and Arg187, and the dominant negative charge is provided by the Glu27 residue at the most solvent-accessible regions. In addition to the small hydrophobic residues, the pocket is also lined by Lys23, Met25, Ser26, Thr122, and Asp173.

• Open in new tab
• Download powerpoint

FIG. 4.

Deep pocket on the protein surface is created by the dimerization of PhnH. In this region, a binding site of acetate is apparent (left inset). The F _o-F _c density at 3σ-level contour is shown for both the acetate ion and a coordinated water molecule. The hydrogen bonds between the acetate ion and the carbonyl oxygens of Pro89 and Ala92 are indicated. There is a high degree of sequence conservation between PhnH from E. coli and homologues from the Pfam database (http://www.sanger.ac.uk/Software/Pfam/ ) for residues lining the pocket (right inset). Select residues that may be important for ligand recognition are indicated.

The flexible loop containing residues 111 to 119 is close to the pocket and creates a binding site for a single acetate ion (Fig. 4). The acetate pocket is created by the side chain atoms of Leu109 and Ala121 and the backbone atoms of Pro89, Ala92, and Ala108. The ion has been oriented during refinement to produce hydrogen bonds between an acetate oxygen atom and the carbonyl oxygen atoms of Pro89 and Ala92. In this orientation, the methyl carbon is positioned toward the backbone atoms of Leu109, resulting in complete solvent exposure of the second oxygen atom of the acetate ion. Electron density was also apparent for a single atom that forms a hydrogen bond with the solvent-exposed oxygen. This density has been modeled as a water molecule, and no additional difference density is observed following refinement of this atom. It is important to note that PhnH was crystallized in the presence of high concentrations of acetate (0.2 M magnesium acetate) at very low pH, and therefore the observed binding may be attributable to the crystallization conditions. However, it is noteworthy that the residues comprising the acetate binding pocket (particularly Leu109) are moderately to strongly conserved in PhnH and are close to the larger pocket formed by the dimer interface. Therefore, the acetate pocket and mechanism of acetate binding may be functionally relevant.

Examination of the sequence conservation of residues in potential members of the PhnH protein family relative to their three-dimensional positions in the E. coli PhnH structure revealed a high degree of sequence identity in the region of this large pocket (Fig. 4). The majority of all the residues strictly conserved among the PhnH homologues are localized within the pocket. These residues include Asp57, Phe94, Thr122, Asp173, Pro186, Arg187, and Thr188. Additionally, Phe18 and Arg19 contribute to the region through dimerization with the second molecule.

PhnH exhibits strong structural similarity to several families of transferases.Automated structural alignment searches using the Protein Structure Comparison Service SSM at the European Bioinformatics Institute (http//www.ebi.ac.uk/msd-srv/ssm ) (15) to search the Structural Classification of Proteins (SCOP) database did not identify a close structural relative of PhnH when the entire tertiary fold was considered the template (1). Some degree of similarity could be observed between PhnH and the highest-ranking structural neighbors with regard to the central β-sheet and flanking helices (Fig. 5). The best matches were 1-aminocyclopropane-1-carboxylate synthase (PDB accession no. 1B8G), RadB recombinase (PDB accession no. 2CVH), alfalfa caffeoyl-CoA 3-O-methyltransferase (PDB accession no. 1SUI), human catechol-O-methyltransferase (PDB accession no. 2AVD), and pig cytosolic aspartate aminotransferase (PDB accession no. 1AJS). The scoring functions employed to identify these structural matches generally provided root mean square deviations of 2.9 to 3.3 Å and calculated z scores of 1.1 to 4.2. It is also important to note that in these analyses, only 89 to 112 residues were structurally matched, while the remaining secondary structures exhibited no measurable alignment. Taken together, the alignment statistics, along with the diversity of global folds and functions of the identified proteins, indicate that PhnH is indeed significantly different from previously determined protein domains.

• Open in new tab
• Download powerpoint

FIG. 5.

Three-dimensional structural alignments of PhnH and proteins whose structures are known, including 1-aminocyclopropane-1-carboxylate synthase (PDB accession no. 1B8G), RadB recombinase (PDB accession no. 2CVH), alfalfa caffeoyl-CoA 3-O-methyltransferase (PDB accession no. 1SUI), human catechol-O-methyltransferase (PDB accession no. 2AVD), and pig cytosolic aspartate aminotransferase (PDB accession no. 1AJS). Automated alignments with members of the PLP-dependent transferase domain yielded reasonable scores for NifS-like protein (PDB accession no. 1EG5) and histidinol-phosphate aminotransferase (PDB accession no. 1UU1). Structural elements exhibiting structural similarity to PhnH are indicated by the same color, and additional elements are gray.

Tertiary structure alignments also suggested some similarity with a variety of nucleotide binding domains. This was, however, judged to be an artifact of the high number of structures exhibiting β-strands sandwiched between alpha layers. Close inspection of the arrangement of secondary structural elements, in particular the ordering of strands within the central β-sheet, revealed that PhnH has a secondary structure arrangement distinct from that of members of the NAD(P)-binding Rossmann fold and the nucleotide binding domain. Additionally, no binding of PhnH to ATP, ADP, AMP, GTP, flavin adenine dinucleotide, flavin mononucleotide, NAD⁺, NADP⁺, 3′-5′-cAMP, adenosine, thiamine diphosphate, S-adenosylmethionine, or acetyl-CoA could be detected using either cocrystallization or isothermal titration calorimetry techniques.

Alternatively, the order of secondary structural elements was manually assessed for structural similarity with protein families described in the SCOP database (1). In this manner, PhnH could be classified as a member of the alpha/beta protein structural superfamily, given the predominantly parallel arrangement of β-strands in the core β-sheet. Analysis of the topology of the large core β-sheet revealed that the component strands are arranged in the order 3245671 with strand 7 antiparallel to the other parallel strands. Along with the alpha/beta global fold, this pattern suggests that the protein is more closely related to the PLP-dependent transferase family (SCOP domain c.67).

PLP is an activated form of vitamin B₆ that serves as a versatile catalyst in numerous enzymatic reactions (8). PLP-dependent enzymes are involved in many cellular processes, most notably in the biosynthesis of amino acids, amino acid-derived metabolites, and amino sugars. These enzymes are generally classified into five distinct structural families, and although the families exhibit poor sequence similarity, each family retains significant structural homology within the core α-β-α fold. No sequence similarity between PhnH and members of this domain family could be detected. Using the SSM server, the PhnH monomer was structurally compared with members of this c.67 SCOP family (Fig. 5). The closest alignments were obtained with a NifS homologue (PDB accession no. 1EG5) and histidinol-phosphate aminotransferase (PDB accession no. 1UU1) from Thermatoga maritima. The z scores were 4.13 and 3.3 and the root mean square deviations were 2.77 and 3.24 Å, respectively. In both cases, the central β-sheet, as well as helices B, C, and E of PhnH, aligned well with the core PLP-dependent transferase domain of each protein. As seen with all alignments of PhnH with various transferases, PhnH is unique in its lack of an extended C-terminal helical domain. As previously noted, the three-strand β-sheet of the mixed alpha/beta layer has also not been identified in other members of this SCOP family. Despite distant similarity to the PLP-dependent transferase fold, we did not observe binding of pyridoxal phosphate to PhnH by UV-visible spectroscopy or isothermal titration calorimetry; likewise, the typical Schiff base interaction between pyridoxal phosphate and PhnH is unlikely as the PhnH family lacks a conserved Lys. Thus, PhnH may represent a protein with a novel function in the PLP-dependent transferase domain family.

Ligand docking trials.The FlexX module of the SYBYL protein modeling suite (Tripos Inc., St. Louis, MO) was employed for virtual docking simulations. Structural coordinates were obtained from the diversity and mechanistic compound libraries available through the National Cancer Institute (http://dtp.nci.nih.gov/ ). Default docking parameters were employed, and the results were scored with CScore within FlexX (2). The entire protein surface was employed as the target docking surface in order to not bias the docking to the putative ligand binding site. While the data did not unambiguously identify a likely ligand or substrate for either the PhnH monomer or dimer, some general trends were observed in the highest-scoring simulations. The best-scoring ligands from both the diversity and mechanistic libraries were all placed within the putative ligand binding cleft that has been previously described (Fig. 6). For brevity, only the top three best-fitting compounds are discussed here. These compounds were NSC 371880 (compound 1), NSC 621486 (compound 2), and NSC 292213 (compound 3). Both compound 1 and compound 2 had a well-aligned phenyl ring situated in the pocket created by Phe18, Leu22, Met25, Leu55, Phe94, Thr122, Asp173, and Arg187. In both molecules this aromatic ring is substituted with a nitro group that is positioned for interaction through the oxygen atoms with the side chain NH₂ atoms of Arg187 and the OD2 atom of Asp173. While interaction with Asp173 by the oxygen residues of the NO₂ group is unlikely, potential hydrogen bonding may be envisioned for compounds in which the NO₂ is replaced by a phosphate moiety. Compound 3 could also be modeled into the pocket of PhnH. This compound likewise exhibits a potential hydrogen bonding interaction between a substituent hydroxyl and these two protein atoms. As Asp173 and Arg187 are conserved residues in the PhnH protein family, the identification of these residues as putative ligand recognition anchors may point to a significant role of these amino acids in the in vivo function of PhnH and its homologues. Also of interest is the conservation of Phe94, which, based on this docking experiment, appears to provide the hydrophobic character necessary for the binding of the substituted aromatic ring.

• Open in new tab
• Download powerpoint

FIG. 6.

Highest-ranking putative ligands, identified through virtual ligand screening and ranked by CScore (1), shown in the lowest-energy conformations. NSC 371880 (compound 1) is green, and NSC 621486 (compound 2) is in orange. Potential interactions between the NO₂ groups of the putative ligands and the Arg187 and Asp173 residues of PhnH are indicated by dashed lines. The conserved Phe94 residue in the lower left portion of the pocket is blue.

While structural similarity searches and virtual ligand docking experiments have not yet revealed a definitive biologically relevant ligand for PhnH, important general trends have emerged. The high proportion of strictly conserved residues found in PhnH homologues located in and around the deep pocket of the PhnH molecular surface suggests that this pocket has significant importance to the in vivo function of PhnH. In particular, the strictly conserved Asp173, Arg187, and Thr188 residues may be especially important in ligand binding and perhaps in catalysis as part of the CP lyase reaction. Additionally, the location of a disordered loop region in the vicinity of the mouth of the pocket indicates that there may be a “lid closure” mechanism for the binding of a biologically relevant ligand. Point mutations of these residues and examination of the competence of the phnH mutants in vivo are necessary to probe the relevance of this conserved pocket.

Activity and binding studies of PhnH with phosphonates.The ability of PhnH to bind several phosphonates and phosphate-containing compounds was assessed by isothermal titration calorimetry and cocrystallization. No binding was detected for methylphosphonic acid, phenylphosphonic acid, propylphosphonic acid, or α-d-ribofuranosyl ethylphosphonate, a compound produced by E. coli during alkylphosphonate metabolism (3). Binding studies were also performed in the presence of calcium and magnesium. These metals, however, failed to promote the interaction of any of these compounds with PhnH. C-P bond cleavage was also not observed by monitoring these compounds by ³¹P-NMR in the presence of PhnH. Recent screening with PhnH for ligand binding and catalytic activity using 430 common metabolites, metals, cofactors, sugars, nucleotides, amino acids, and antibiotics also proved to be unsuccessful (A. Yakunin, personal communication). This result suggests that the PhnH pocket described above, if it is relevant to the in vivo function of PhnH, is specific for an as-yet-unidentified substrate or ligand. It is important to also consider the possibility that the enzymatically competent state of PhnH may be achieved only when PhnH interacts with another member of the phn operon to form the active holoenzyme or the possibility that PhnH has a structural or scaffold role in the CP lyase complex rather than a direct catalytic role.

Expression of phnH is essential to CP lyase activity in E. coli.Mutants defective in individual genes of the phn operon were analyzed for the production of methane from degradation of methylphosphonate. The strains also contained the pstS605 allele, which results in constitutive expression of the members of the pho regulon. Table 4 indicates that the wild-type strain produced appreciable amounts of methane, whereas none of the strains containing phnG to phnM mutations produced amounts of methane above the background level. Similarly, a strain with the entire phn operon deleted was unable to produce methane. The strains defective in phnE, which are defective in phosphonate transport, or the strains defective in phnF produced methane at a level above background but lower than that of the wild-type strain. For comparison, we also qualitatively analyzed the ability of the phn mutant strains to grow on methylphosphonate (Table 1). In agreement with previous growth studies of Metcalf and Wanner (17, 18) and consistent with the methane production data described above, none of the phn mutant strains grew with methylphosphonate as the phosphorus source. We therefore conclude that the phnH gene product is essential for C-P bond cleavage activity.

Conclusions.Structural analysis of PhnH reveals that it is marginally similar to the PLP-dependent transferase domain; however, the inability of PhnH to bind PLP, the absence of the C-terminal domain, and the presence of novel secondary structures compared to members of the c.67 SCOP family strongly suggest that PhnH and its homologues represent a novel, functionally distinct subfamily of this domain. Although virtual ligand screening and experimental trials have not yet identified an in vivo ligand or substrate for PhnH, these studies have suggested that PhnH may be capable of binding aromatic molecules through interaction of polar substituents with strictly conserved residues in a putative binding pocket. While this study highlights the continuing challenges posed by in vitro characterization of CP lyase, it establishes that PhnH is an essential component of the C-P bond-cleaving activity of CP lyase and represents the first structural footing with which to probe the mechanism of this remarkable enzyme reaction.