Three-dimensional structure of xylonolactonase from Caulobacter crescentus: A mononuclear iron enzyme of the 6-bladed β-propeller hydrolase family.

Xylonolactonase Cc XylC from Caulobacter crescentus catalyzes the hydrolysis of the intramolecular ester bond of d-xylonolactone. We have determined crystal structures of Cc XylC in complex with d-xylonolactone isomer analogues d-xylopyranose and (r)-(+)-4-hydroxy-2-pyrrolidinone at high resolution. Cc XylC has a 6-bladed β-propeller architecture, which contains a central open channel having the active site at one end. According to our previous native mass spectrometry studies, Cc XylC is able to specifically bind Fe2+ . The crystal structures, presented here, revealed an active site bound metal ion with an octahedral binding geometry. The side chains of three amino acid residues, Glu18, Asn146, and Asp196, which participate in binding of metal ion are located in the same plane. The solved complex structures allowed suggesting a reaction mechanism for intramolecular ester bond hydrolysis in which the major contribution for catalysis arises from the carbonyl oxygen coordination of the xylonolactone substrate to the Fe2+ . The structure of Cc XylC was compared with eight other ester hydrolases of the β-propeller hydrolase family. The previously published crystal structures of other β-propeller hydrolases contain either Ca2+ , Mg2+ , or Zn2+ and show clear similarities in ligand and metal ion binding geometries to that of Cc XylC. It would be interesting to reinvestigate the metal binding specificity of these enzymes and clarify whether they are also able to use Fe2+ as a catalytic metal. This could further expand our understanding of utilization of Fe2+ not only in oxidative enzymes but also in hydrolases.

asymmetric unit

Caulobacter crescentus

Diamond Light Source

European Synchrotron Radiation Facility

(r)‐(+)‐4‐hydroxy‐2‐pyrrolidinone

paraoxonase

senescence marker protein 30

xylose dehydrogenase

xylonolactonase

xylonate dehydratase

INTRODUCTION

Pentoses, the five‐carbon sugars (in particular d‐xylose and l‐arabinose), are abundant in nature and can be found in large amounts in biomasses such as grasses, crops, and hardwoods and are consequently promising raw materials for the production of bio‐based chemicals. The investigation of microbial metabolic pathways has revealed a number of enzymes, which can participate in the processing and modifying of pentose sugars. The oxidative pathway, called Dahms pathway, is utilized to convert an abundant pentose carbohydrate, d‐xylose, to pyruvate and glycolaldehyde. In the first step, d‐xylose is oxidized to d‐xylonolactone by d‐xylose dehydrogenase. In the second step of the pathway, d‐xylonolactone is hydrolyzed to d‐xylonic acid.

d‐Xylonolactone is also prone to being slowly hydrolyzed non‐enzymatically to d‐xylonic acid, but it has been reported that xylonolactonase improves the d‐xylonolactone hydrolysis and thus production of d‐xylonic acid both in in vivo and in vitro pathway studies. , , Furthermore, in bacteria, the genes coding for the Dahms pathway enzymes are often arranged in one operon, which further suggests that the lactonase is needed for an efficient conversion of the xylonolactone to xylonic acid. d‐Xylonolactone exists as two isomers, d‐xylono‐1,5‐lactone (d‐xylono‐δ‐lactone) and d‐xylono‐1,4‐lactone (d‐xylono‐γ‐lactone). The two isomers of d‐xylonolactone are likely to interconvert via a bicyclic intermediate described in References 5, 6, 7. Only d‐xylono‐1,4‐lactone is available commercially, which may indicate difficulties in isolation of d‐xylono‐1,5‐lactone.

Xylonolactonase (E.C. 3.1.1.68) catalyzes the hydrolysis of the intramolecular ester bond of d‐xylonolactone (Figure 1a). The lactonase Cc XylC gene is found in C. crescentus bacterium in the same operon as the Cc XylB dehydrogenase and Cc XylD dehydratase enzymes. Thus, it can be assumed that the real substrate for Cc XylC is d‐xylonolactone. In solution, d‐xylose exists almost completely either in α‐d‐xylopyranose (34%) or β‐d‐xylopyranose (64%) diastereoisomeric forms. The oxidation of the pyranose form results in d‐xylono‐1,5‐lactone, which is probably the substrate for Cc XylC. This is also supported by the observation that the structurally very similar d‐glucono‐1,5‐lactone is also a substrate for Cc XylC.

FIGURE 1

(a) The hydrolysis of d‐xylonolactone, which is suggested to involve the isomerization reaction of d‐xylono‐1,4‐lactone to d‐xylono‐1,5‐lactone as the first step, which is catalyzed non‐enzymatically by a divalent metal ion (M2+). The lactonase‐catalyzed hydrolysis reaction produces d‐xylonic acid, which has relatively low pK a (see also Figure 5). (b) The ligands successfully used as cryoprotectants and observed in the active site of the enzyme

FIGURE 5

The proposed reaction mechanism for the Cc XylC catalyzed hydrolysis of d‐xylono‐1,5‐lactone to form d‐xylonic acid. See the text for more details

We have recently discussed that divalent metal ions (such as Ca2+, Zn2+, and Fe2+) are able to promote the non‐enzymatic isomerization of d‐xylono‐1,4‐lactone to d‐xylono‐1,5‐lactone and thus improve the enzymatic hydrolysis to xylonic acid (Figure 1a). , The hydrolysis of its ester bond produces d‐xylonic acid, which has a rather low pK a (about 3.65), which leads to acidification of the medium. The decrease of pH has often been used in in vitro lactonase assays to monitor enzymatic activity.

So far, detailed enzymatic characterization of xylonolactonases has been very limited. We have previously studied the Cc XylC, which was found on the basis of amino acid sequence homology to have a 6‐bladed β‐propeller fold. , , In addition, we have recently found that Cc XylC unexpectedly binds to a single Fe2+ ion with high specificity and affinity and not, for example, to Ca2+ or Zn2+. Therefore, we have continued the analysis of Cc XylC by determining the three‐dimensional structure of the enzyme to elucidate the binding of Fe2+ and its structural role in catalysis.

RESULTS AND DISCUSSION

Overall structure

We have determined three high‐resolution crystal structures of the Cc XylC xylonolactonase (Table 1). Two structures were obtained in the presence of d‐xylopyranose and represent two crystal forms, which have differences in ligand binding outside the active site. The third structure was obtained with (r)‐(+)‐4‐hydroxy‐2‐pyrrolidinone ((r)‐β‐hydroxy‐γ‐butyrolactam, HPD, Figure 1b). The expressed polypeptide chain of the Cc XylC contained 290 amino acid residues. The mass spectrum analysis had previously shown that the initial methionine was cleaved, and it was consequently absent in the crystallized protein material. The overall electron density maps were of good quality but there are weaker densities for the four first residues of the Cc XylC in one molecule in the asymmetric unit, indicating partial disorder in the N‐terminus. The three‐dimensional structures show that the Cc XylC is folded into a β‐propeller protein having six blades, that is, repeats of four‐stranded antiparallel β‐sheets. The blades are arranged around the central channel, which contains the active site at one end. The structure of the Cc XylC is rather regular without any larger insertions except one protruding loop with a short α‐helix in blade 1. In addition, there is a 13‐residue‐long C‐terminal polypeptide tail, which packs against the propeller and forms a cap over the central channel at the opposite end to the active site (Figure 2a).

TABLE 1

X‐ray data collection and refinement statistics

PDB ID	7PLB	7PLC	7PLD
Crystal form	Rectangle	Parallelogram	Rectangle
Ligand	d‐Xylopyranose	d‐Xylopyranose	(r)‐(+)‐4‐Hydroxy‐2‐pyrrolidinone
Beamline	DLS i04‐1	DLS i04‐1	DLS i04‐1
Wavelength (Å)	0.91587	0.91587	0.91587
Space group	C 2 2 21	P 1 21 1	C 2 2 21
a, b, c (Å)	85.8, 171.0, 79.1	45.8, 82.2, 159.2	87.7, 171.7, 79.2
α, β, γ (°)	90, 90, 90	90, 97.7, 90	90, 90, 90
Resolution (Å)	85.58–1.73 (1.79–1.73)	158.24–2.15 (2.23–2.15)	85.97–1.70 (1.76–1.70)
Observations	121,864 (11,959)	115,401 (11,849)	131,593 (12,969)
Unique observations	61,016 (6,025)	59,670 (6,048)	65,883 (6,493)
Completeness	100.0% (99.9%)	93.5% (95.2%)	99.7% (98.9%)
I/σ I	9.2 (1.1)	3.9 (1.0)	9.2 (1.0)
CC1/2	0.998 (0.564)	0.991 (0.660)	0.999 (0.538)
R merge	0.049 (0.719)	0.100 (0.551)	0.048 (0.784)
R meas	0.069 (1.017)	0.141 (0.779)	0.068 (1.109)
R work	0.1807	0.1978	0.1936
R free	0.2178	0.2502	0.2343
Number of protein monomers in the asymmetric unit	2	4	2
Number of protein atoms	4,504	8,829	4,457
Number of waters	553	611	579
Number of ligand atoms	147	89	86
Ramachandran favorable (%)	97.73	97.71	97.37
Ramachandran outliers (%)	0	0	0
Rms deviation in bond lengths (Å)	0.003	0.003	0.006
Rms deviation in bond angles (°)	0.692	0.630	0.815
Average B factor (Å2)	28.34	28.58	27.30

FIGURE 2

(a) The overall three‐dimensional structure of the Cc XylC. The polypeptide chain of monomeric protein is presented as a green cartoon model. The blades of the β‐propeller are numbered from 1 to 6. The bound Fe2+ ion in the central channel is shown as a brown ball and d‐xylopyranose as a stick model. (b) The superimposed eight monomers from the three crystal structures are shown as grey ribbon chains. The additional bound ligand molecules are shown as green stick models

Crystal structures

The asymmetric units (asu) of the three Cc XylC crystal structures contained eight crystallographically independent monomers (2/asu + 4/asu + 2/asu) in total (Table 1). All of the structures had a strong electron density peak in the active site and were modelled as an Fe2+ ion according to our earlier native mass spectrometric studies. The presence of the bound iron was also verified with native mass spectrometry from the protein crystals (Figure S1).

Two substrate analogues were observed in the complex structures. The molecular structure of d‐xylopyranose (d‐xylose) is very similar to that of d‐xylono‐1,5‐lactone except that the pyranose ring has a hydroxyl substituent instead of a carbonyl group, and both ligands have a six‐membered ring. Another substrate analogue used was a cyclic amide, HPD, which has a five‐membered ring and an amide group thus representing an analogue of d‐xylono‐1,4‐lactone. The two crystal structures complexed with d‐xylopyranose had clear electron densities for bound d‐xylopyranose (Figure S2A) in the active site in all six molecules (2/asu + 4/asu) of the asymmetric units. The d‐xylopyranose molecules were bound as β‐anomers in all enzyme structures in a very similar fashion. The equatorial O1 is bound to the Fe2+ ion with a short bond distance (2.0 Å) indicating strong binding to the metal cation. The binding can be considered as stereospecific for β‐anomer and not for α‐anomer, which has the O1 in an axial position, which would not allow the bond formation with Fe2+. β‐d‐Xylopyranose also forms several hydrogen bonds with Cc XylC protein. O1 and O2 form hydrogen bonds with Asn101, O2 with Asn101 and Arg99, O3 and O4 with Glu120, and O5 with Asn146. In addition, Trp211 has a clear role in binding since it packs against the flat ring of d‐xylopyranose (Figure 3a).

FIGURE 3

The active site and bound ligands. The protein structures are shown as green stick models and ligand structures as cyan stick models. Iron atoms and water oxygens are presented as balls in brown and red, respectively. The lengths of hydrogen bonds and coordination bonds are given in ångströms. The side chains from three amino acid residues (Glu18, Asn146, and Asp196), which are bound to the Fe2+, are all in the same plane (planar triad). (a) The bound d‐xylopyranose in the structure of the rectangular crystal, molecule A. (b) The bound HPD in the crystal structure, molecule B. (c) The active site without the ligand in the crystal structure with HPD, molecule A. (d) Modelled binding of d‐xylono‐1,5‐lactone (in yellow) to the active site (same as in c). (e) Modelled binding of d‐xylono‐1,4‐lactone (in yellow) to the active site (same as in c). (f) All three experimental structures superimposed together with the models of two isomers of d‐xylonolactone (in yellow)

The crystal structure obtained with HPD showed electron densities (Figure S2B) for four ligand molecules in one Cc XylC molecule (chain B) and one in the other (chain A). One of the ligands was bound to the active site in the same site as d‐xylopyranose and showed clear electron density that corresponds to a planar carbonyl group and atoms in α‐positions. The rest of the molecule was disordered, probably because of multiple binding modes. The binding of the HPD was slightly tilted when compared to the binding of d‐xylopyranose. The carbonyl oxygen was strongly bound to Fe2+ (distance 1.9 Å). There are also possibilities for hydrogen bonds between the HPD and Asn101 or Asp196 depending on the binding mode (Figure 3b). We modelled the binding mode of d‐xylono‐1,4‐lactone to the active site by mimicking the binding modes of d‐xylopyranose and HPD (Figure 3e) and found several close contacts between the atoms of d‐xylono‐1,4‐lactone and the side chains of the amino acid residues in the active site. d‐Xylono‐1,4‐lactone has the C5 in an axial position, which renders its structure different to d‐xylono‐1,5‐lactone, which has all substituents in equatorial positions. In consequence, it seems that Cc XylC prefers d‐xylono‐1,5‐lactone as substrate over d‐xylono‐1,4‐lactone.

In addition to the ligands bound to the active site, electron densities for ligands were observed in nine additional binding sites on the Cc XylC surface (Figure 2b). In most cases there was only one observed ligand binding among eight protein molecules. In the HPD complex structure, there was a second HPD molecule near the Fe2+‐bound HPD molecule. This second molecule packs against the flat face of bicyclic indole part of Trp211 (Figure 3b). This further supports the role of the side chain of Trp211 as binding to the flat face of cyclic carbohydrate molecules.

We found unexpectedly that one of the additional binding sites had electron densities for a ligand in four protein molecules. HPD binding was observed once in this site, and α‐d‐xylopyranose binding was observed in three structures. This binding site is located between blades 2 and 3 (Figure 2b), and it forms a clear pocket on the Cc XylC surface, although less deep compared to the active site pocket (Figure 4a, b). α‐d‐Xylopyranose forms hydrogen bonds from O3 to His23 of blade 2, and from O4 to Gly67 of blade 3 (Figure 4c). According to the crystal structures, the binding of α‐d‐xylopyranose seems to be stereospecific as we did not observe any density for O1 in the equatorial position, which would exist in the β‐anomer of d‐xylopyranose. It is possible that Trp28 from blade 2 restricts the binding of β‐d‐xylopyranose. There are significant alterations in the loop conformations upon ligand binding (Figure 4c). It is thus possible that this additional binding site would be an allosteric site for α‐d‐xylopyranose participating, for example, in regulation of the oxidative d‐xylose pathway. However, we do not have any other experimental evidence supporting this suggestion, and it would also be challenging to study because d‐xylopyranose exists in solution as a mixture of α‐ and β‐anomers in which β‐anomer binds to the active site and α‐anomer to the surface pocket.

FIGURE 4

The binding of d‐xylopyranose anomers to the Cc XylC. Ligands are shown as cyan stick models and protein surfaces in green. (a) β‐d‐Xylopyranose binding to the active site. (b) α‐d‐Xylopyranose binding to the putative allosteric site on the protein surface. (c) Details of the putative allosteric site. The protein atoms in complex with α‐d‐xylopyranose are in green, and without the ligand in grey

Metal coordination

All the eight Cc XylC molecules in the crystal structures have a similar structural architecture for Fe2+ binding. There is a mononuclear Fe2+ center that is coordinated to six ligand atoms having an octahedral geometry. Fe2+ is coordinated to the side chains of Glu18 (blade 2), Asn146 (blade 5), and Asp196 (blade 6). There are three vacant sites, which are all occupied in the observed crystal structures. Two vacant sites are always occupied by water molecules, and the third site can be occupied by an oxygen of a substrate or, when a substrate or other ligand is not bound to the active site, by a water molecule (Figure 3). The structure without the bound ligand has three Fe2+‐coordinated water molecules. Two of them are strongly bound (distance 2.1 Å), whereas the third one has a longer distance (2.5 Å). The two strongly bound water molecules are probably the two water molecules, which were detected earlier in the native mass spectrum of the Cc XylC in the presence of Fe2+ and also in the native mass spectrum by using a sample made of dissolved crystals (Figure S1).

The coordination of Fe2+ in the Cc XylC resembles to some extent the coordination of iron in other non‐heme enzymes, such as mononuclear non‐heme iron oxygenases. These enzymes contain the so‐called 2‐His‐1‐carboxylate facial triad, where the Fe2+ is coordinated to three amino acid side chains, thus leaving three vacant positions for solvent or substrate atoms, similarly to the Cc XylC. However, in the facial triad of the oxygenases, the three coordinating amino acids are located on one face of the metal cluster whereas in the Cc XylC structure the three coordinating amino acids are all in the same plane around the iron; the coordination of which could thus be called a “planar triad” (Figure 3). The differences in the binding geometries probably reflect different enzymatic activities. The enzymes having the facial triad are all oxidative enzymes capable of activating dioxygen, whereas the Cc XylC is a hydrolase.

Reaction mechanism

According to the crystal structure, β‐d‐xylopyranose binds to the active site in a very defined manner in which O1 is coordinated to the catalytic Fe2+. β‐d‐Xylopyranose can be considered as a competitive inhibitor and a substrate analogue. It is very well superimposable with the molecular model of d‐xylono‐1,5‐lactone, thus providing a view on how the actual substrate would bind (Figure 3f). On the basis of the complex structures, we present a reaction mechanism in which the substrate, d‐xylono‐1,5‐lactone, first binds to the Fe2+ (Figure 5). This increases the electrophilicity of the carbonyl carbon C1, so that a weak nucleophilic water can attack C1. The formed intermediate is then deprotonated, followed by the breaking of the ester bond, and finally the product, d‐xylonic acid, is released. The proton exchange results in the final product, d‐xylonate anion (Figures 1a and 5). The proposed reaction mechanism includes only one catalytically relevant interaction with the enzyme, the binding to Fe2+. In fact, the crystal structures do not suggest any other putative amino acid residues, which would, for example, promote a nucleophilic attack of water to the carbonyl carbon. According to the literature, divalent metal ions alone are able, at least weakly, to catalyze the ester bond hydrolysis. , , , Consequently, it might be possible that by the formation of a substrate–enzyme complex, the lifetime of otherwise weak association of Fe2+ and the carbonyl oxygen increases considerably in appropriate orientations. This in turn would allow a weak nucleophile, a water molecule, to attack and initiate the intramolecular ester bond hydrolysis reaction.

6‐Bladed β‐propeller hydrolase family

When homologous proteins were searched in the Protein Data Bank, we found in total eight enzymes, which showed amino acid sequence identities with the Cc XylC (Table 2). They all have a similar 6‐bladed β‐propeller fold, are hydrolases, and have a catalytic mononuclear metal ion (Figure 6a). The metal ion in the solved crystal structures is either calcium, magnesium, or zinc but not iron. The coordination of metal is similar among these enzymes (Figure 6b). A detailed comparison is presented in Figure 6c,d, which shows the metal binding in Cc XylC and mouse SMP30. When considering the metal coordination, all of the eight enzymes have coordinating amino acids in the same plane, but there are either three or four amino acids, which participate in metal ion binding (Figure 6b).

TABLE 2

Representative crystal structures of 6‐bladed β‐propeller hydrolase family

Enzyme	Source	PDB	Resolution (Å)	Amino acids	Identity to cc XylC %	Active site metal	Metal coordinating amino acids	Ligand (substrate analogue)
Xylonolactonase	Caulobacter crescentus	7PLB	1.7	289	100	Fe2+	E18, N146, D196	d‐Xylopyranose
SMP30	Mouse	4GN8	1.7	299	30	Ca2+	E18, N154, D204	1,5‐Anhydro‐d‐glucitol
SMP30	Human	4GNC	1.8	299	29	Ca2+	E18, N154, D204, D104	1,5‐Anhydro‐d‐glucitol
Luciferin‐regenerating enzyme	Photinus pyralis	5D9B	1.5	311	30	Mg2+	E18, N160, D212	No
Gluconolactonase	Xanthomonas campestris	3DR2	1.7	305	19	Ca2+	E48, N134, N191, D242	No
Drp35 gluconolactonase	Staphylococcus aureus	2DG1	1.7	333	14	Ca2+	E48, D138, N185, D236, S237	No
Diisopropylfluoro‐phosphatase	Loligo vulgaris	2GVW	1.7	314	14	Ca2+	E21, N120, N175, D229	Dicyclopentyl phosphoramidate
Paraoxonase (PON)	Oryctolagus cuniculus	1V04	2.2	355	11	Ca2+	E53, N168, N224, D269	No

FIGURE 6

Superimposition of representative members (Table 2) of the 6‐bladed β‐propeller hydrolase family. (a) Cartoon representation of protein monomers. The Cc XylC is in green, other proteins are in grey. The active site ligands are shown as stick models. (b) The superimposition of metal coordination in active sites. The proteins that have three amino acids in plane, which participate in metal ion binding, are in green (xylonolactonase, mouse and human SMP30, and luciferin‐regenerating enzyme), and proteins that have four amino acids in plane are in cyan (Drp35, gluconolactonase, diisopropylfluorophosphatase, and paraoxonase). The interacting residues in the Cc XylC are labelled as E18, D196, and N146. (c,d) The detailed comparison of coordination of catalytic metal ion in Cc XylC (c), and mouse SMP30 (d). The bond lengths between the metal ion and ligand atoms are given in ångströms. Carbon atoms of proteins are in green and of bound ligand (xylose in C, and 1,5‐anhydro‐d‐glucitol in d) in cyan. Both crystal structures have been determined at 1.7 Å resolution. The comparison shows a clear resemblance in metal binding

Three enzymes have about 30% amino acid identities with the Cc XylC, namely mouse and human senescence marker proteins (SMP30) and Photinus pyralis luciferin‐regenerating protein. All these enzymes have three amino acids residues, which are bound to the metal cation (Table 2). The crystal structures have been determined for human and mouse SMP30. The biological role of SMP30 proteins has been vague, but it has been reported that they are able to hydrolyze various substrates such as d/l‐glucono‐δ‐lactone, d/l‐gulono‐γ‐lactone and d/l‐galactono‐γ‐lactone and also to catalyze lactone‐ring formation. The crystal structures for human and mouse SMP30 have been determined with several ligands such as 1,5‐anhydro‐d‐glucitol, which is bound approximately to the same position as the ligands in the crystal structures of the Cc XylC. The metal ion is coordinating to Glu18 (blade 2), Asn154 (blade 5), and Asp204 (blade 6), in a very similar manner as the Fe2+ in the Cc XylC. Another enzyme, which has three similar metal‐coordinating residues (Glu18, Asn160, and Asp212), is the luciferin‐regenerating enzyme, but its biological function is unclear. This crystal structure does not contain any bound ligand in the active site. The protein was crystallized in presence of MgCl2, and in consequence, the bound metal ion was interpreted as Mg2+ (Figure 6b).

Four other 6‐bladed β‐propeller enzymes have a lower sequence identity with Cc XylC (Table 2), and they all have four amino acid residues participating in metal ion binding (Figure 6b). The crystal structure of the gluconolactonase from Xanthomonas campestris has been determined at high resolution. The active site does not include bound ligand but contains Ca2+, which is coordinated to four amino acid residues. Three of these coordinating amino acids can also be found in similar positions as in the Cc XylC, namely Glu48 (blade 2), Asn134 (blade 4), Asp242 (blade 6), and the fourth is an additional one, Asn191 (from blade 5). The corresponding asparagine (Asn101) in the Cc XylC structure is located further away from the Fe2+ (4.2 Å) to be able to coordinate to the iron. The crystal structure of Staphylococcus aureus Drp35 includes Ca2+ and similar coordinating residues Glu48, Asn185, and Asp236 as the Cc XylC does. Mutating any of these residues to alanine has been reported to result in an almost inactive enzyme. Mutations Asn185Asp and Ser237Ala, on the other hand, led only to reduced activity, indicating that these non‐conserved residues are not so critical for metal ion binding. S. aureus Drp35 has been suggested to belong to paraoxonases, which act as lactonases but are also able to inactivate various organophosphates, including insecticides and nerve gases. Diisopropyl fluorophosphatase from Loligo vulgaris is also able to hydrolyze organophosphorous nerve agents. Its crystal structure complexed with dicyclopentylphosphoramidate shows a coordination to Ca2+ by the oxygen of the phosphoramidate group of the substrate analogue. The metal ion has a similar coordination to four amino acid residues as in the previously mentioned X. campestris gluconolactonase. Finally, the crystal structure of rabbit paraoxonase PON1 variant obtained by directed evolution has been determined to have a phosphate ion in the active site. The coordination of the Ca2+ ion is mediated by similar four coordinating amino acid residues as described above.

CONCLUSIONS

To our knowledge, the Cc XylC represents the first example of a hydrolytic enzyme having Fe2+ as a catalytic metal ion. As deduced from the complex structures, the major contribution to the ester bond hydrolysis is due to the coordination of carbonyl oxygen of the lactone ring to the enzyme‐bound Fe2+, which could promote a water molecule to attack the carbonyl carbon. The comparisons of the three‐dimensional structure of the Cc XylC with other hydrolases from the 6‐bladed β‐propeller family showed that the metal ion coordination is highly conserved. This suggests that other members of this enzyme family could also utilize Fe2+ as a catalytic metal. However, verifying this would require further experimental studies. 6‐Bladed β‐propeller hydrolases, in general, show activity towards a wide range of substrates, such as esters, lactones, and phosphate esters. If the essential feature in catalysis requires only the coordination of the oxygen atom of a substrate to a divalent metal ion, there are more possibilities for variations in the active site structure. This would allow the utilization of members of this enzyme family to catalyze hydrolytic reactions with versatile substrates.

MATERIALS AND METHODS

Crystallization

The Cc XylC encoding gene was cloned into the linearized (NcoI and XhoI digestion) Escherichia coli pBAT4 expression vector. The N‐terminal methionine codon was followed by an additional alanine codon, added to create a NcoI site. The expressed polypeptide chain of the Cc XylC thus contained 290 amino acid residues. The Cc XylC was produced and purified as a monomeric protein in a single anionic exchange chromatography step, as described earlier and crystallized by using hanging drop vapor diffusion. For crystallization, approximately 20 M equivalents of FeSO4 was added to the protein, and the solution was desalted with a PD‐10 desalting column (GE Healthcare, Little Chalfont, England) and concentrated to about 8 mg/ml in 50 mM Tris buffer (pH 8.0). Drops were made by combining 1 μl protein and 1 μl precipitant solutions over 500‐μl reservoirs, and crystallization plates were kept at 20°C after preparation. Initial screening was done with Crystal Screen by Hampton Research (HR2‐110), and the best hits were obtained with high ammonium or lithium sulfate concentrations. Small, disordered crystals were obtained with 1.6 M (NH4)2SO4 and 0.1 M HEPES (pH 7.5) as the precipitant solution. Streak seeding with a dog hair from these crystals at 1.6–1.8 M (NH4)2SO4 yielded very small needle‐like crystals. Another streak seeding or alternatively microseeding (drop diluted 1/320 with the precipitant solution used in drop preparation) from these microcrystals at 1.5 M (NH4)2SO4 yielded clean two‐dimensional rectangular crystals with rough maximum dimensions 200 × 200 × 10 μm. Another crystal was also observed: when 2.0% (vol/vol) PEG400 and 0.05 M sodium malonate were used in addition to the 1.6 M (NH4)2SO4 and 0.1 M HEPES (pH 7.5), clusters of parallelogram‐shaped crystals were obtained. Streak seeding from these crystals in the same conditions yielded clean crystals with similar size as the rectangular crystals.

X‐ray crystallography

The crystals were transferred to cryoprotective solutions and stored in liquid nitrogen for X‐ray crystallography measurements. The (NH4)2SO4 concentrations used in crystallization were too high for the cryoprotective solution because it would precipitate when cooled down to 100 K, so 1.0 M (NH4)2SO4 was used along with 0.1 M HEPES (pH 7.5) and 30% (wt/vol) cryoprotectant. In addition, with the parallelogram‐shaped crystals, 2.0% (vol/vol) PEG400 was added as well. The crystals were unstable in the presence of most cryoprotectants, which were tried, including ethylene glycol, glycerol, and d‐xylonolactone. However, they were stable with d‐xylopyranose and HPD, which were conveniently also substrate analogues. The crystals were picked up in nylon loops, stored in liquid nitrogen and sent to be measured at the synchrotrons at European Synchrotron Radiation Facility (ESRF) and Diamond Light Source (DLS). All herein presented data were obtained at the i04‐1 beamline at DLS, and an early dataset that resulted in the first preliminary structure was measured at the ID30A‐3 (MASSIF‐3) beamline at ESRF.

Structure determination

The first preliminary structure was calculated using the early dataset. The images were processed to 2.8 Å resolution with XDS program package, and using this data file, calculations were done with PHENIX software suite. Using the homologous mouse SMP30 gluconolactonase (PDB entry 4GN7) as the initial model, molecular replacement was done with MR‐Rosetta and Phaser. Extensive manual editing of the model with Coot and structure refinement with phenix.refine resulted in a model, which contained few water molecules and lacked any ligands but could be used as an initial model for future calculations.

For determining the high‐resolution structures, datasets of three crystals were selected: rectangular with d‐xylopyranose, parallelogram‐shaped with d‐xylopyranose, and rectangular with HPD. Autoprocessed data files from DLS with the best statistics – xia2‐dials, , , , , autoPROC , , , , , and autoPROC for the three datasets respectively—were selected and used in calculations. Molecular replacement was done with Phaser using the preliminary structure as the starting model, and structure refinement was done with phenix.refine. Geometry restraints were calculated for the HPD using eLBOW. Water molecules were initially modelled automatically using the “Update waters” option and later examined manually. Fe2+ ions and identifiable ligands were also added manually and refined automatically, and their occupancies were also refined if they were significantly less than one. Weight optimization parameters were enabled for the final refinement rounds. To further verify the presence of HPD in the active site of the corresponding structure, a polder map was also calculated, and it showed the expected strong intensity at most of the lactam ring (Figure S3).

Mass spectrometry

To show that the bound iron was present in crystalline protein, a native mass spectrometry sample was made from the rectangular crystals. They were picked up in a nylon loop and transferred to fresh drops of crystallization solution where they were dissolved by adding water. After enough crystals were collected, the drops were combined and buffer‐exchanged to 10 mM ammonium acetate using a PD‐10 desalting column. Using a Solarix XR Fourier transform ion cyclotron resonance (FT‐ICR) mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany), a native mass spectrum was measured with the same parameters as in the previous work but with 300 scans. A clean mass spectrum with signals of both apo‐ and holo‐forms of the enzyme was obtained, the holo‐form being dominant (Figure S1). When another sample was prepared similarly but by dissolving the crystals in 10 mM FeSO4 solution before the buffer exchange and the spectrum was measured, the weak apo‐signals disappeared and the holo‐signals remained, as expected.

CONFLICT OF INTEREST

The authors declare that they have no competing interests.

AUTHOR CONTRIBUTIONS

Johan Pääkkönen: Formal analysis (equal); investigation (equal); validation (equal); visualization (equal); writing – original draft (equal). Nina Hakulinen: Funding acquisition (equal); investigation (equal); resources (equal); writing – review and editing (equal). Martina Andberg: Investigation (equal); writing – review and editing (equal). Anu Koivula: Conceptualization (equal); funding acquisition (equal); investigation (equal); resources (equal); writing – review and editing (equal). Juha Rouvinen: Conceptualization (equal); funding acquisition (equal); investigation (equal); project administration (equal); supervision (equal); visualization (equal); writing – review and editing (equal).

Data S1 Native mass spectrum of protein crystals, electron density maps, polder map.

Figure S1. (a) Charge state deconvoluted native mass spectrum of Cc XylC crystals showing that the crystalline protein is mostly (more than 90%) in holo‐form with bound Fe2+. (b) The original native mass spectrum of Cc XylC crystals.

Figure S2. The representative electron densities for (a) d‐xylopyranose and (b) HPD (two binding modes, the ligands rotated by roughly 180° along the carbonyl bond). 2F −F electron density maps are in blue (1 sigma level). F −F difference maps are in red (negative 3 sigma level) and in green (positive 3 sigma level).

Figure S3. The polder map of HPD ligand, contour level 5σ. Two binding modes for HPD are presented on top of each other.

Click here for additional data file.