Bacterial utilization of pectin as a carbon source is an important metabolic feature of free-living soil bacteria responsible for soft-rot diseases in plants.1 Extracellular unsaturated oligogalacturonides resulting from the breakdown of pectin are further metabolized intracellularly to a common intermediate, 2-keto-3-deoxygluconate, which can ultimately be converted to pyruvate. The enzyme, 5-keto-4-deoxyuronate isomerase (KduI), is a component of this pathway and catalyzes the interconversion of 5-keto-4-deoxyuronate and 2,5-diketo-3-dexoygluconate. KduI (EC 220.127.116.11) belongs to a class of enzymes containing the well-studied isomerases triose phosphate isomerase and xylose isomerase, both of which exhibit characteristic (β/α)8- or TIM-barrel folds. Not all members of this EC class are TIM-barrels and, as KduI shares little sequence homology with other members of its class, it was of interest to determine the structure of KduI. Here we present the crystal structure of KduI from Escherichia coli determined by multiwavelength anomalous dispersion (MAD) and show that the enzyme binds metal and is structurally homologous to members of the cupin superfamily.
Materials and Methods.
Cloning, expression, and purification of KduI:
DNA was prepared from the E. coli strain DH5α with 5′- and 3′-oligonucleotides designed for PCR cloning into the NdeI/BamHI sites of pET15b. The cloned gene contained the expected sequence with the exception of the C-terminal 5′ residues. From Swiss-Prot Q46938, the expected amino acid sequence at the C-terminus of KduI is K275-D276-L277-R278, while our observed sequence is K275-E276-I277-C278-A279. Cloned KduI was transferred into the strain B348 for expression and purification of selenomethionine-substituted protein. Bacteria were grown at 37°C in minimal medium supplemented with selenomethionine to an OD600 of 0.6. After induction with 0.2 mM IPTG the temperature was lowered to 16°C and incubation continued for an additional 20–24 h while shaking at 250 rpm. Cells were harvested by centrifugation at 6000 × g for 20 min. Crude KduI was prepared by chelation chromatography on Ni-NTA-Sepharose and further purified by ion exchange chromatography over Q-Sepharose FF in 50 mM HEPES, 1 mM DTT (pH 7.5). Following removal of the His6-tag by thrombin digestion (0.5 U/mg of KduI, 20°C, 12–18 h) KduI was subjected to final purification by ion exchange over Q-Sepharose FF. The protein was further characterized by SDS-PAGE and by dynamic light scattering using a DynaPro-801 Molecular Sizing Instrument from Protein Solutions, Inc.
KduI was concentrated to 10–15 mg/mL protein and frozen in 40–50 μL aliquots at −80°C until used. Aliquots were thawed and dialyzed against 10 mM HEPES, 25 mM NaCl, 2 mM DTT (pH 7.5) prior to crystallization. Crystals of selenomethionine-substituted KduI were grown by vapor diffusion at 4°C using drops of 3:2 mixtures of protein with reservoir solutions of 20–30% PEG 400 containing 0.2 M CaCl2 and 100 mM HEPES (pH 7.0). Crystals were frozen in liquid N2 following a brief exposure to a cryosolution of 30% PEG 400, 100 mM HEPES, 0.2 M CaCl2, 1 mM DTT (pH 7.0) and 10% glycerol. Selenomethionine-substituted KduI crystals were grown from PEG 400 in space group R3 (a = b = 102.9 Å, c = 176.9 Å, hexagonal setting) with two molecules of KduI in the asymmetric unit (Matthews coefficient of 2.9 with an estimated 56% solvent content). This crystal form of KduI did not suffer from the apparent twinning problems observed for native KduI crystals grown from salt solutions.2
Data collection, structure determination, and refinement:
A 3-wavelength MAD dataset was collected on a single crystal of KduI to 1.94 Å on beamline X8C at the NSLS, Brookhaven National Laboratories, Upton, NY. Data were integrated and scaled using HKL2000.3 The structure of KduI was solved using the autoSHARP4 interface to the CCP4,5 SHARP,6 and ARP/wARP7 routines. Rantan found 24 sites (24 Se sites were expected; each monomer of KduI possesses 12 methionines) that were subsequently refined in SHARP. The initial electron density map obtained from SHARP was solvent-flattened in DM,5, 8 and the flattened map was used to build a model with ARP/wARP. The initial map was of sufficient quality to allow ARP/wARP to build almost 2/3 of the entire structure without intervention. The remainder of the model was built manually using the XBuild package of Quanta (Accelrys, Inc.). Further refinement of the model was accomplished by simulated annealing and/or conjugant-gradient minimization using the CNX suite of programs (Accelrys, Inc.). Electron density maps and models were displayed and intermediate models were modified in Quanta. Merging and refinement statistics for KduI are shown in Table I. Figures were prepared using Molscript9 and Raster3D.10 Electron density surfaces were prepared using CONSCRIPT.11 Coordinates and structure factors have been deposited with the PDB under accession id 1XRU.
Table I. Data Collection and Refinement Statistics
Numbers in parentheses are for the highest resolution shell.
Rsym = ΣΣi|Ii − 〈I〉|/ΣΣiIi where I is the integrated intensity of a reflection.
Rfactor = Σhkl∥Fobs − kFcalc|/Σhkl|Fobs|. 5% of all reflections were omitted from refinement and Rfree is the same statistic calculated for these reflections.
The structure of a single KduI subunit consists of two similar halves comprised of 12 β-strands. There are five α-helices distributed throughout the two β-barrels and one short stretch of 310 helix (η1) along the extended chain connecting the two halves of the molecule. Except for the short parallel β1/β3 packing, all the strands are arranged in antiparallel sheets as shown in Figure 1(A). A hexamer is generated by the crystallographic symmetry applied to the two subunits in the asymmetric unit [see Fig. 1(B)]. This finding is consistent with a previous gel filtration study in which KduI was found to be a hexamer in solution2 as well as dynamic light scattering measurements in which the molecular weight of KduI was estimated to be 165 kDa corresponding to five to six molecules of the 31-kDa KduI. Each KduI subunit has two surfaces of approximately 3000 Å2 that serve as interfaces between adjacent molecules of the hexamer; one, of 3192 Å, forms the AB′ and another, of 2975 Å, forms the AB interfaces as shown in Figure 1(B). Though relatively short and few in number, the α-helices are involved in unique pairing interactions characteristic of the interfaces of the hexamer. The α3 helices of the A and B molecules interact along the AB subunit interface while the α4 and α5 helices interact with their symmetry-related counterparts across the AB′ interface. The α1 and α2 helices line the central hole of the hexamer [Fig. 1(B)].
KduI is a metalloprotein:
Electron density maps of KduI reveal a bound metal ion in the C-terminal half of the molecule. The metal is liganded by three histidines (His196, His198, and His245) and a glutamate (Glu203), potentially representing a catalytic site in the enzyme. The geometry of the liganding is tetrahedral with an average bond angle between the metal and liganding atoms of 109.3°. Bond lengths between the metal and its ligands range from 1.9–2.2 Å. A water molecule 3.7 Å away could be weakly coordinated, but is not obviously liganded to the metal.
Calcium ions were present during crystallization, but the bound metal is not likely to be calcium as it is rarely liganded to nitrogens.12, 13 Other potential candidates are zinc, nickel, manganese, iron, or copper. We suggest that the bound metal is likely a zinc atom. The atom must have a measurable anomalous signal within the MAD dataset, since the initial solution of the selenium atom substructure placed 24 atoms, two of which turn out to be extra metal atoms, while the remaining sites correspond to 22 of the expected 24 selenium atoms. At the wavelengths used for the experiment, zinc is expected to have the greatest anomalous signal of the likely candidates for the bound metal. When the metal site is refined as zinc, the B-factor of the bound atom correlates well with the B-factors of the liganding atoms, whereas if the metal is refined as manganese the B-factor of the bound atom is roughly half that of the liganding atoms. The coordination of the bound metal is tetragonal with a coordination sphere of four ligands. Square-planar coordination is more commonly observed for copper- and iron-containing compounds whereas tetrahedral coordination is more common for compounds containing zinc.12, 14 We cannot absolutely rule out nickel, iron or copper, but we tentatively identify the bound metal as zinc.
An ordered PEG molecule is near K165 and the metal ion:
We noticed an unusual feature of the electron density maps in the region surrounding Lys165. There appears to be an open ring of unexplained density surrounding the Nζ atom. The identity of the bound substance is not known; however, the size and shape of the density suggests that it might be a molecule of polyethelyne glycol, which was used as the precipitant for crystallization. We were able to model the density as five units of ethylene glycol [HOCH2(OCH2CH2O)4CH2OH] as shown in Figure 1(C). A molecular replacement solution of the crystal structure of KduI crystallized from ammonium sulfate does not have the extra density near Lys165 (Dr. Pete Dunten, personal communication) further indicating that the bound substance is most likely polyethylene glycol.
KduI from Erwinia chrysanthemi has been characterized and shown to be a sugar isomerase.1, 15 Based on the similarity between the E. coli and Erwinia KduI sequences, which are 75% identical, we would predict that the E. coli KduI is also a sugar isomerase and might be expected to bind ring-opened or ring-closed carbohydrate molecules. It could be of interest that the bound PEG molecule is close (ca. 7 Å) to the zinc atom and 5.4 Å away from one of the carboxylate oxygens of Glu203 that ligates the bound metal. In addition, there is a water molecule situated midway between the bound metal (3.70 Å Zn-to-water distance) and one of the atoms of the polyethylene glycol moiety (3.72 Å water-to-carbon atom distance), which may mediate chemistries with molecules bound in the vicinity of K165. We note that K165 is not absolutely conserved among KduI genes; it is often replaced by a glutamine, for example. Such a substitution could still position a potential proton donor/acceptor in the vicinity of the putative binding site, but we do not know if the PEG binding we observe here is a general finding of all KduIs. It is possible that the polyhydroxy compound, PEG, mimics the binding mode of either substrate or product of KduI.
KduI from E. coli is structurally homologous to members of the cupin family:
We were unable to identify any sequences homologous to KduI for which a structure had been previously determined. A number of homologous genes were identified by a PSI-BLAST search but no structures of these proteins had been reported (Fig. 2). As we prepared this manuscript for publication we became aware of a PDB deposition (1X8M) reporting the structure of KduI from E. coli. At the outset of our structure determination, however, there were no suitable templates for use as models for molecular replacement, which led us to do the MAD experiment. The protein fold seen in 1X8M appears to be the same as that which we present here, but we observe the additional features of metal and PEG binding. To our knowledge, the 1XM8 structure has not yet been reported in the literature. With the crystal structure of KduI in hand, it became apparent that this protein fold has been seen before. A DALI search16, 17 of known structures found a number of proteins with folds similar to that of KduI (Table II). Twelve of the top fourteen structures found by the DALI search belong to the cupin family of proteins (two are unclassified hypothetical proteins). The fold of KduI from E. coli is clearly very similar to that of the cupin family of proteins.18–20 Yet this protein failed to be selected by searches for cupin-homologous sequences. Cupin-related proteins exhibit a characteristic bipartite sequence homology20 with motif 1 [G(X)5HXH(X)3,4E(X)6G] separated from motif 2 [G(X)5PXG(X)2HXN] by a variable length stretch of 15–50 intervening residues. KduI has some sequence similarity to motif 1 in the region of the metal binding site around H196–E203, and the separation between this cluster of residues and the final metal-coordinating residue H245 is within the 15–50 residue separation of the consensus cupin motifs. The fit to the consensus for motif 2 is much less obvious in the region of H245. The sequence identity between KduI and the various cupin family members ranges from 3–16%. Thus, although KduI does not include the conserved sequence motifs found in other cupins, based on its structural similarity to other family cupin members, we suggest that it is in fact a new member of the cupin family.
Table II. DALI Alignment of Proteins Structurally Similar to KduI
Number of atoms
E. coli hypothetical protein
Plant glycinin g1
B. subtilis oxalate decarboxylase
Quercetin 2,3-dioxygenase (bacterial)
Plant glycinin g5
Yeast phosphomannose isomerase
B. subtilis phosphomannose isomerase
Human homogentisate 1,2-dioxygenase
T. martima hypothetical protein
P. furiosus 6-phosphoglucose isomerase
Plant oxalate oxidase (germin)
We thank Dr. Pete Dunten for his encouragement and helpful discussions, and members of the Georgiadis lab for helpful discussions and assistance.