A New Role for the Ankyrin Repeat Revealed by the Study of the N-formyltransferase from Providencia alcalifaciens

Abstract

N -formylated sugars such as 3,6-dideoxy-3-formamido-D-glucose (Qui3NFo) have been observed on the lipopolysaccharides of various pathogenic bacteria including Providencia alcalifaciens, a known cause of gastroenteritis. These unusual carbohydrates are synthesized in vivo as dTDP-linked sugars. The biosynthetic pathway for the production of dTDP-Qui3NFo requires five enzymes with the last step catalyzed by an N-formyltransferase that utilizes N¹⁰-tetrahydrofolate as a cofactor. Here we describe a structural and functional investigation of the P. alcalifaciens N-formyltransferase, hereafter referred to as QdtF. For this analysis the structure of the dimeric enzyme was solved in the presence of N⁵-formyltetrahydrofolate, a stable cofactor, and dTDP-3,6-dideoxy-3-amino-D-glucose (dTDP-Qui3N) to 1.5-Å resolution. The overall fold of the subunit consists of three regions with the N-terminal and middle motifs followed by an ankyrin repeat domain. Whereas the ankyrin repeat is a common eukaryotic motif involved in protein:protein interactions, reports of its presence in prokaryotic enzymes have been limited. Unexpectedly, this ankyrin repeat houses a second binding pocket for dTDP-Qui3N, which is characterized by extensive interactions between the protein and the ligand. To address the effects of this second binding site on catalysis, a site-directed mutant protein, W305A, was constructed. Kinetic analyses demonstrated that the catalytic activity of the W305A variant was reduced by approximately sevenfold. The structure of the W305A mutant protein in complex with N⁵-formyltetrahydrofolate and dTDP-Qui3N was subsequently determined to 1.5-Å resolution. The electron density map clearly showed that ligand binding had been completely abolished in the auxiliary pocket. The wild-type enzyme was also tested for activity against dTDP-3,6-dideoxy-3-amino-D-galactose (dTDP-Fuc3N) as a substrate. Strikingly, sigmoidal kinetics were observed indicating homotropic allosteric behavior. Although the identity of the ligand that regulates QdtF activity in vivo is presently unknown, our results still provide the first example of an ankyrin repeat functioning in small molecule binding.

The lipopolysaccharide or LPS is a complex glycoconjugate associated with the outer membranes of Gram-negative bacteria where it provides a permeability barrier to hydrophobic or negatively charged molecules.¹ It can be envisioned in terms of three specific components: the lipid A, the core oligosaccharide, and the O-specific polysaccharide or O-antigen. It is the O-antigen, which extends farthest away from the bacterium that contributes to the wide species variations seen in nature. Oftentimes quite unusual dideoxysugars are found on the O-antigens of pathogenic Gram-negative bacteria including 3,6-dideoxy-3-formamido-D-glucose (Qui3NFo) or 3,6-dideoxy-3-formamido-D-galactose (Fuc3NFo), which are shown in Scheme 1a. The first step in the biosynthetic pathway for the production of Qui3NFo in bacteria involves the attachment of a dTMP moiety to glucose-1-phosphate to yield dTDP-glucose. This step is followed by a 4,6-dehydration, a 3,4-ketoisomerization, and an amination reaction. The last step in the biosynthetic pathway is catalyzed by an N-formyltransferase, which requires N¹⁰-formyltetrahydrofolate for activity (Scheme 1b).

Scheme 1.

Herein we report a structural and functional investigation on the N-formyltransferase from Providencia alcalifaciens, a pathogenic organism linked to the 1996 food poisoning outbreak in Japan.² The high-resolution structures of the enzyme, hereafter referred to as QdtF, were determined in the presence of N⁵-formyltetrahydrofolate (a stable analog, Scheme 1c) and dTDP-sugar substrates. Strikingly, the N-terminal catalytic region of the enzyme is followed by ankyrin repeats. The occurrence of such repeats in QdtF is highly unusual as they are typically found in eukaryotic proteins. First identified in the Drosophila signaling protein Notch and in the yeast cell cycle regulator Swi6/Cdc10, the ankyrin repeat is now known to be one of the most commonly occurring molecular architectures in eukaryotic proteins involved in such processes as the cell cycle, transcriptional regulation, signal transduction, and modulation of the inflammatory response.^3–6

Remarkably, the ankyrin repeat domain in QdtF harbors a second binding pocket for the dTDP-sugar substrate that is characterized by extensive interactions between the ligand and protein side chains. To the best of our knowledge, ankyrin repeats have never been implicated in small molecule binding. Taken together the molecular and biochemical data from our investigation expand the repertoire of ankyrin repeats from protein:protein interactions to small molecule binding and allosteric control.

Materials and Methods

Protein Constructs, Expression, and Purification

The gene encoding QdtF from P. alcalifaciens serogroup O40 was synthesized by DNA2.0 with optimized Escherichia coli codon usage. An N-terminal tag of the sequence MGSSHHHHHHSSENLYFQGA was included before the initial methionine residue, and this construct was placed into the pJ411 vector for protein expression. Details concerning protein purification are provided in Supporting Information. The protein was dialyzed against 10 mM Tris-HCl (pH 8.0) and 200 mM NaCl and concentrated to 8 – 12 mg/mL based on an extinction coefficient of 0.87 (mg/mL)⁻¹cm⁻¹ as calculated with DNASTAR.

Crystallization of QdtF

Prior to crystallization trials, the protein was incubated with 5 mM dTDP and 5 mM N⁵-formyltetrahydrofolate (N⁵-formyl-THF). Crystallization conditions were then surveyed by the hanging drop method of vapor diffusion using a sparse matrix screen developed in the laboratory. Single crystals of QdtF were grown by hanging drop diffusion against a precipitant solution containing 100 mM HEPPS (pH 8.0), 200 mM KCl, and 5 – 7% poly(ethylene glycol) 8000. The crystals belonged to the space group P4₁2₁2 with unit cell dimensions of a = b = 78.3 Å and c = 151.1 Å. The asymmetric unit contained one subunit.

The dTDP-sugar complexes were obtained by soaking the crystals in a solution composed of 100 mM HEPPS (pH 8.0), 15% poly(ethylene glycol) 8000, 200 mM KCl, 200 mM NaCl, 5 mM N⁵-formyl-THF, and either 10 mM dTDP-Qui3N or dTDP-Fuc3N. The solution was exchanged 3 – 4 times to ensure removal of dTDP. The dTDP-Qui3N or dTDP-Fuc3N substrates were prepared as described in Supporting Information.

Site-Directed Mutagenesis and Crystallization

The site-directed mutant protein, W305A, was generated via the QuikChange method of Stratagene. The protein variant was expressed and purified as described for the wild-type enzyme. Crystals were grown and soaked under conditions similar to that for the wild-type enzyme.

Structural Analysis of QdtF

Prior to X-ray data collection, the crystals of the wild-type enzyme/dTDP-Qui3N/N⁵-formyl-THF complex were transferred to a cryo-protectant solution containing 100 mM HEPPS (pH 8.0), 25% poly(ethylene glycol) 8000, 18% ethylene glycol, 300 mM KCl, 300 mM NaCl, 5 mM N⁵-formyl-THF, and 10 mM dTDP-Qui3N. X-ray data were collected at 100K with a Bruker AXS Platinum 135 CCD detector controlled by the Proteum software suite (Bruker AXS Inc.). The X-ray source was Cu Kα radiation from a Rigaku RU200 X-ray generator equipped with Montel optics and operated at 50 kV and 90 mA. These X-ray data were processed with SAINT version 7.06A (Bruker AXS Inc.) and internally scaled with SADABS version 2005/1 (Bruker AXS Inc.). Relevant X-ray data collection statistics are listed in Table S1 of Supporting Information. The structure was solved via molecular replacement with the software package PHASER using as a search model the coordinates for another sugar N-formyltransferase determined in the laboratory (PDB code 4LXQ).^7,8 Iterative cycles of model-building with COOT and refinement with REFMAC reduced the R_work and R_free to 18.5% and 21.5%, respectively, from 30 - 1.5 Å resolution.^9,10 X-ray data from the wild-type enzyme/dTDP-Fuc3N/N⁵-formyl-THF complex crystals were collected and processed in a similar manner. This complex structure was determined via Difference Fourier methods. Iterative cycles of model building with COOT and refinement with REFMAC reduced the R_work and R_free to 20.7% and 26.5%, respectively, from 30 - 1.8 Å resolution.

Crystals of the W305A variant with dTDP-Qui3N and N⁵-formyl-THF were prepared and X-ray data collected in a similar manner to that described above. Relevant X-ray data collection statistics are provided in Table S1 of Supporting Information. Iterative cycles of model building with COOT and refinement with REFMAC reduced the R_work and R_free to 18.1% and 20.8%, respectively, from 30 - 1.50 Å resolution.^9,10 Model refinement statistics for all three protein complexes are provided in Table S2 of Supporting Information.

Kinetic Analysis

In general, the kinetic constants for wild-type QdtF with its natural substrate dTDP-Qui3N or with the alternative substrate dTDP-Fuc3N or the W305A variant of QdtF with dTDP-Qui3N or dTDP-Fuc3N were determined via a discontinuous assay using an ÄKTA HPLC. The reaction rates were determined by calculating the amount of the formylated product (dTDP-Qui3NFo or dTDP-Fuc3NFo) produced on the basis of the peak area on the HPLC trace as measured at 267 nm. The area was correlated to concentration via a calibration curve created with standard samples that had been treated in the same manner as the reaction aliquots.

Specifically, the kinetic parameters for the wild-type enzyme with dTDP-Qui3N as a substrate were determined using 1 mL reactions containing 5 mM N¹⁰-formyl-THF, 50 mM HEPPS (pH 8.0), 0.83 μM enzyme, and substrate concentrations ranging from 0.05 – 4.0 mM dTDP-Qui3N. Four 250-μL aliquots were taken at time points over four minutes and quenched by the addition of 12 μL of 6 M HCL. Afterwards, 200 μL of carbon tetrachloride were added, the samples were vigorously mixed, spun at 14000 × g for 1 minute, and 200 μL of the aqueous phase taken for subsequent analysis via HPLC. The samples were diluted with 2 mL water and loaded onto a 1 mL ResQ column. The products were quantified after elution with an eight-column volume gradient of 0 – 400 mM LiCl (pH 4, HCl). The kinetic parameters of the wild-type enzyme with dTDP-Fuc3N as a substrate were determined in a similar manner using 1.25 μM enzyme and 0.5 – 20 mM dTDP-Fuc3N.

The kinetic parameters for the W305A variant with dTDP-Qui3N as a substrate were determined using 2.5 μM enzyme and 0.2 – 18 mM dTDP-Qui3N. Finally, the kinetic parameters for the W305A variant with dTDP-Fuc3N as a substrate were determined using 5.2 μM enzyme and 0.40 – 50.0 mM dTDP-Fuc3N. Plots of concentrations versus initial rates were analyzed using PRISM (GraphPad Software, Inc.) and were fitted to the equation v_o = (V_max[S])/(K_M + [S]) for hyperbolic data and v_o = (V_max[S]^h)/((K_M)^h + [S]^h) for sigmoidal data. Relevant kinetic parameters are listed in Table 1.

Table 1.

Kinetic Parameters.

Enzyme	Km (mM)	kcat (s−1)	kcat/Km (M−1 s−1)
wild-type with dTDP-Qui3N	0.45 ± 0.09	2.43 ± 0.31	5.4 × 103
W305A variant with dTDP-Qui3N	2.15 ± 0.35	1.75 ± 0.15	8.0 × 102
wild-type with dTDP-Fuc3N	4.18 ± 0.43	1.50 ± 0.21	3.6 × 102
W305A variant with dTDP-Fuc3N	5.52 ± 0.72	1.82 ± 0.32	3.3 × 102

Results and Discussion

The first structure of QdtF determined for this investigation utilized crystals of the enzyme in complex with its substrate, dTDP-3,6-dideoxy-3-amino-D-glucose (dTDP-Qui3N), and N⁵-formyltetrahydrofolate (N⁵-formyl-THF), a stable but catalytically inert cofactor. The model was refined to a nominal resolution of 1.5 Å (R_overall = 18.5%, R_free = 21.5%). Overall, the quality of the electron density was excellent such that it was possible to trace the entire polypeptide chain from Met 1 to Asn 397. The enzyme has one cis peptide linkage at Pro 99, which is located at ~14 Å from the active site. The histidine tag used for protein purification was not visible in the electron density map.

QdtF, like other sugar N-formyltransferases, functions as a dimer.^11,12 In the crystals utilized for this investigation (P4₁2₁2 space group), QdtF packed with its local twofold rotational axis coincident to the crystallographic dyad, thereby reducing the contents of the asymmetric unit to one subunit. A ribbon representation of the subunit is displayed in Figure 1a. As can be seen, it is composed of the three domains: the N-terminal region that is dominated by eight β-strands and six helices (Met 1 to Leu 223), the middle motif that is composed of a β-hairpin and a four-stranded antiparallel β-sheet (Pro 224 to Asn 271), and the C-terminal region, which contains ankyrin repeats (Phe 272 to Asn 397). The ankyrin repeat itself is a helix-loop-helix motif composed of ~33 amino acid residues with the α-helices running antiparallel and an extended loop at the C-terminus projecting outwards at an ~90° angle.⁶ Most proteins with ankyrin motifs contain between four to seven repeats but some, such as the cytoskeletal protein ankyrin, contain 24.^13,14 In the case of QdtF, there are three complete ankyrin repeats.

Figure 1.

Overall structure of QdtF. The enzyme crystallized in the space group P4₁2₁2 with one subunit in the asymmetric unit. A ribbon representation of the subunit is displayed in stereo in (a). The N-terminal, middle, and C-terminal domains are highlighted in blue, cyan, and purple, respectively. The dTDP-sugar ligands and the N⁵-formyl-THF cofactor are depicted in stick representations. The quaternary structure of QdtF is dimeric as shown in (b). A surface representation of the dimer is depicted in stereo in (c). Note that both the active site and the allosteric binding pocket are quite shallow and solvent exposed.

A ribbon representation of the QdtF dimer, with overall dimensions of ~66 Å × 66 Å × 130 Å, is shown in Figure 1b. The total buried surface area at the dimer interface is ~2100 Ų. Notably, whereas the N-terminal and middle domains of each subunit provide side chains for the formation of the subunit:subunit interface, the ankyrin repeats splay away from the main body of the molecule. The N-terminal domain harbors the active site cleft, which is shallow and solvent accessible (Figure 1c). The ankyrin repeat region provides a second shallow binding pocket for dTDP-3,6-dideoxy-3-amino-D-glucose. A distance of ~25 Å separates the two dTDP-sugar ligand binding pockets within the subunit.

Shown in Figures 2a and 2b are the observed electron densities corresponding to the two dTDP-sugar ligands observed binding to the QdtF subunit. The dTDP-Qui3N substrate adopts an extended conformation in the active site, whereas in the auxiliary binding pocket, it assumes a quite curved structure with the ribose C-3 hydroxyl oxygen lying within 3.0 Å of the hexose endocyclic oxygen. A close-up view of the QdtF active site is provided in Figure 3a. The thymine ring of dTDP-Qui3N is wedged between Tyr 195 and Tyr 221 and is hydrogen bonded to three waters molecules and the carboxamide group of Gln 222. The backbone amide nitrogens of Tyr 105 and Ser 107, and the carbonyl oxygens of Phe 76 and Gly 103 provide additional protein:substrate interactions. Twelve water molecules surround the dTDP-Qui3N substrate. With the exception of Gln 222, however, there are no other side chains that directly interact with the dTDP-sugar. The positions of Asn 92, His 94, and Asp 130 are shown in Figure 3a since these residues are conserved amongst various N-formyltransferases including WlaRD from Campylobacter jejuni,¹¹ ArnA from E. coli,¹⁵ E. coli L-methionyl-tRNA N-formyltransferase,¹⁶ and E. coli glycinamide ribonucleotide transformylase.¹⁷ These amino acid residues have been implicated in catalysis. In the case of WlaRD, for example, the N94A, H96N, and D132N site-directed mutant proteins demonstrate no enzymatic activity under the assay conditions employed.¹¹

Figure 2.

Observed electron density for the bound dTDP-sugar ligands. The electron density corresponding to the dTDP-sugar bound in the active site is displayed in (a). The map, contoured at 3σ was calculated with coefficients of the form F_o - F_c, where F_o was the native structure factor amplitude and F_c was the calculated structure factor amplitude. The electron density corresponding to the dTDP-sugar bound in the allosteric binding pocket is shown in (b). The map was calculated and contoured as described in (a). Interactions within 3.2 Å between the dTDP-sugars and the protein side chains are indicated by the dash lines.

Figure 3.

Close up views of the active site and the allosteric binding pocket. Shown in (a) is the active site region surrounding the dTDP-sugar substrate. Protein:ligand interactions ≤3.2 Å are indicated by the dashed lines. Ordered water molecules are depicted as red spheres. The dTDP-sugar substrate and N⁵-formyl-THF are colored in gray bonds, whereas the protein bonds are displayed in white. Shown in (b) is the allosteric binding pocket with the protein side chains colored in purple to emphasize that they belong to the ankyrin repeat domain. As in (a), protein:ligand interactions ≤3.2 Å are indicated by the dashed lines, and ordered solvents are depicted as red spheres. The green dashed line indicates the presence of an intramolecular hydrogen bond in the dTDP-sugar substrate when it is bound in the allosteric binding pocket.

A close-up view of the auxiliary binding pocket is shown in Figure 3b. In contrast to that observed in the active site, in the auxiliary binding pocket there are numerous interactions between the dTDP-sugar and the protein side chains. Specifically Trp 305, Tyr 313, Asn 334, Lys 336, Thr 338, Tyr 343, and Asp 346 serve to anchor the dTDP-Qui3N ligand into the ankyrin repeat domain. To address the role of the auxiliary binding pocket in the catalytic activity of QdtF, a site-directed mutant protein, W305A, was subsequently prepared, its enzymatic activity measured, and its structure solved to 1.5-Å resolution in the presence of dTDP-Qui3N and N⁵-formyl-THF (R_overall = 18.1%, R_free = 20.8%). It was reasoned that removal of the parallel stacking interaction between the indole ring of Trp 305 and the thymine ring of dTDP-Qui3N would abolish dTDP-sugar binding. Indeed, the electron density map showed no trace of dTDP-sugar binding in the second pocket, whereas the electron density corresponding to dTDP-Qui3N in the active site was unambiguous (Figure S1, Supporting Information). Importantly, the α-carbons for the wild-type enzyme and the W305A mutant protein superimpose with a root-mean-square deviation of 0.12 Å, thus indicating that removal of ligand binding to the site located in the ankyrin domain had little effect on the overall three-dimensional structure of QdtF. Ordered solvent molecules simply moved into the position vacated by the dTDP-Qui3N ligand.

As described in Materials and Methods, the catalytic activities of the wild-type enzyme and the W305A protein were investigated via an HPLC discontinuous assay. Plots of initial velocities versus dTDP-Qui3N concentrations for the wild-type enzyme and the W305A variant demonstrated Michaelis-Menten kinetics (Figure S2, Supporting Information). Relevant kinetic parameters are listed in Table 1. The catalytic efficiency of the W305A protein is reduced from 5400 M⁻¹s⁻¹ to 800 M⁻¹s⁻¹. Whereas the k_cat values for the two proteins are similar within experimental error, removal of ligand binding from the auxiliary pocket increased the K_m for dTDP-Qui3N from 0.45 mM for the wild-type enzyme to 2.15 mM for the W305A variant. Clearly, the binding of a dTDP-sugar to the auxiliary pocket of QdtF plays a subtle yet measureable allosteric role. Whether the binding of the dTDP-sugar to the auxiliary pocket affects the rate of formation of the Michaelis complex or the rate of its dissociation to free enzyme and substrate is not known, however.

Given the observation that the active site region surrounding the C-4′ amino group of dTDP-Qui3N is relatively open, the activity of QdtF against dTDP-3,6-dideoxy-3-amino-D-galactose (dTDP-Fuc3N) was also tested. QdtF is, indeed, capable of turning over the alternative substrate albeit at a reduced catalytic efficiency of 360 M⁻¹s⁻¹ (Table 1). Strikingly, an initial velocity plot using the wild-type enzyme against dTDP-Fuc3N concentration demonstrated sigmoidal kinetics (Figure 4a). The calculated Hill slope of 5.1 ± 0.6 indicates positive cooperativity. These data demonstrate that QdtF exhibits homotropic allostery with respect to the dTDP-Fuc3N substrate and thus has the proper architecture for communication between the active site and the auxiliary binding pocket. The identity of the ligand that regulates the activity of QdtF in vivo is not presently known. As a control, the catalytic activity of the W305A variant was also tested against the alternative substrate dTDP-Fuc3N. Hyperbolic kinetics were again observed as would be expected given the removal of ligand binding from the ankyrin domain (Table 1 and Figure 4b).

Figure 4.

Kinetic behavior using dTDP-Fuc3N as a substrate. A plot of the reaction rate versus concentration of dTDP-Fuc3N for the wild-type enzyme is shown in (a). Panel (b) shows the plot for the W305A mutant protein.

To complete the investigation of QdtF, its structure was determined in the presence of dTDP-Fuc3N and N⁵-formyl-THF to 1.8-Å resolution. The electron densities corresponding to the dTDP-Fuc3N ligand in the active site and in the auxiliary binding pocket, respectively, are shown in Figures 5a and 5b. The α-carbons for the QdtF models with either bound dTDP-Qui3N or dTDP-Fuc3N superimpose with a root-mean-square deviation of 0.13 Å. The active site geometries of QdtF in the presence of either dTDP-Qui3N or dTDP-Fuc3N are nearly identical within experimental error. The C-3′ amino group is shifted by ~0.4 Å in the active site. With respect to the allosteric pocket, the only major change occurs in the side chain conformation of Asp 346. When dTDP-Qui3N is bound, the carboxylate group of Asp 346 lies within hydrogen bonding distance to the C-4′ hydroxyl group of the pyranosyl ring (Figure 3b). In the presence of dTDP-Fuc3N, the side chain of Asp 346 swings out of the pocket.

Figure 5.

Electron density for the dTDP-Fuc3N ligand bound to wild-type QdtF. The observed electron density for the dTDP-Fuc3N ligand in the active site is shown in (a). The map was calculated as described in Figure legend 2 and contoured at 3σ. The N⁵-formyl-THF cofactor is highlighted in pink bonds. The observed electron density for the dTDP-Fuc3N sugar in the auxiliary binding site is displayed in (b). The map was contoured at 3σ.

The majority of ankyrin repeat proteins studied thus far have been isolated from eukaryotes, although it is becoming apparent from amino acid sequence analyses that the motif is also found in some bacterial and viral proteins.^4,14,18,19 Little is known, however, regarding the physiological roles of ankyrin repeats in prokaryotes. Their prevalence in proteins from intracellular pathogens such as Legionella pneumophila is especially intriguing, however.²⁰ A search of the QdtF sequence against the National Center for Biotechnology data bank shows that the enzyme is homologous to an additional eight putative N-formyltransferases that contain ankyrin repeats (Figure S3, Supporting Information). Importantly, Trp 305 and Tyr 313 in QdtF are completely conserved amongst these proteins suggesting that the allosteric binding pocket is functional in these enzymes as well. Many of these putative N-formyltransferases are found in pathogenic organisms such as Pantoea stewartii, which causes Stewart’s wilt of sweet corn, Francisella tularensis, the causative agent of rabbit fever, and Pseudomonas fuscovaginae that is responsible for brown sheath rot in rice. The discovery of an ankyrin repeat domain in QdtF that harbors an additional dTDP-binding site was completely unexpected, and indeed, ankyrin repeats have never been implicated in small molecule binding. This study thus paves the way for new investigations focusing on the use of ankyrin repeats in allosteric regulation.

Abbreviations

dTDP

thymidine diphosphate

dTMP

thymidine monophosphate

HEPPS

N-2-hydroxyethylpiperazine-N′-3-propanesulfonic acid

HPLC

high-performance liquid chromatography

Tris

tris-(hydroxymethyl)aminomethane