Crystal Structure of Papaya Glutaminyl Cyclase, an Archetype for Plant and Bacterial Glutaminyl Cyclases

doi:10.1016/j.jmb.2005.12.029

J. Mol. Biol. (2006) 357, 457–470

Crystal Structure of Papaya Glutaminyl Cyclase, an Archetype for Plant and Bacterial Glutaminyl Cyclases Rene´ Wintjens1†, Hassan Belrhali2†, Bernard Clantin3 Mohamed Azarkan4, Coralie Bompard3, Danielle Baeyens-Volant4 Yvan Looze1 and Vincent Villeret3* 1

Laboratoire de Chimie Ge´ne´rale Institut de Pharmacie-U.L.B. CP 206/04, Boulevard du Triomphe, B-1050 Brussels Belgium 2

EMBL-Grenoble outstation 6 rue Jules Horowitz, BP F-38042 Grenoble cedex 9 B-1050 France 3

CNRS-UMR 8161, Institut de Biologie de Lille, BP 477181 1, rue du Professeur Calmette F-59021 Lille, France 4

Laboratoire de Chimie Ge´ne´rale I, Faculte´ de Me´decine-U.L.B. CP609, 808, route de Lennik B-1070 Brussels, Belgium

Glutaminyl cyclases (QCs) (EC 2.3.2.5) catalyze the intramolecular cyclization of protein N-terminal glutamine residues into pyroglutamic acid with the concomitant liberation of ammonia. QCs may be classified in two groups containing, respectively, the mammalian enzymes, and the enzymes from plants, bacteria, and parasites. The crystal structure of the ˚ QC from the latex of Carica papaya (PQC) has been determined at 1.7 A resolution. The structure was solved by the single wavelength anomalous diffraction technique using sulfur and zinc as anomalous scatterers. The enzyme folds into a five-bladed b-propeller, with two additional a-helices and one b hairpin. The propeller closure is achieved via an original molecular velcro, which links the last two blades into a large eight stranded b-sheet. The zinc ion present in the PQC is bound via an octahedral coordination into an elongated cavity located along the pseudo 5-fold axis of the b-propeller fold. This zinc ion presumably plays a structural role and may contribute to the exceptional stability of PQC, along with an extended hydrophobic packing, the absence of long loops, the three-joint molecular velcro and the overall folding itself. Multiple sequence alignments combined with structural analyses have allowed us to tentatively locate the active site, which is filled in the crystal structure either by a Tris molecule or an acetate ion. These analyses are further supported by the experimental evidence that Tris is a competitive inhibitor of PQC. The active site is located at the C-terminal entrance of the PQC central tunnel. W83, W110, W169, Q24, E69, N155, K225, F22 and F67 are highly conserved residues in the C-terminal entrance, and their putative role in catalysis is discussed. The PQC structure is representative of the plants, bacterial and parasite enzymes and contrasts with that of mammalian enzymes, that may possibly share a conserved scaffold of the bacterial aminopeptidase. q 2005 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: glutamine cyclotransferase; pyroglutamic acid; X-ray structure; b-propeller; metalloprotein

Introduction Glutaminyl cyclases (QCs) (EC 2.3.2.5) catalyze the conversion of the N-terminal glutaminyl residue of proteins into pyroglutamic acid (5-oxopro† R.W. and H.B. contributed equally to this work. Abbreviations used: QC, glutaminyl cyclase; PQC, papaya glutaminyl cyclase; SAD, single-wavelength anomalous dispersion. E-mail address of the corresponding author: [email protected]

line) with the concomitant liberation of ammonia as shown in Scheme 1. The existence of such an enzyme was first discovered in the latex of the tropical species Carica papaya, primarily known for being a rich source of cysteine proteinases.1–3 Sequencing projects have thereafter revealed the presence of QCs not only in several other plants,4 but also in bacteria and human parasites. The usefulness of such an enzymatic activity may seem intriguing, since glutamine is well known to spontaneously convert into pyroglutamic acid. However, under physio-

0022-2836/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.

458

Papaya Glutaminyl Cyclase Crystal Structure

O

O

NH

NH H2N

QCs

HN

H2NOC

O NH3

Scheme 1.

logical conditions this cyclization is rather slow5 and thereby may require an enzymatic catalysis. Proteins with a similar enzymatic activity have also been purified from mammalian tissues.6,7 These mammalian enzymes however showed no sequence homology with plant QCs, even if both enzymes are monomeric proteins and have close molecular masses.4,8,9 Human QC was predicted to adopt an a/b topology and the enzyme exhibits a remarkable instability above pH 8.5 and below pH 6.0.9–11 The papaya QC (PQC), by contrast, was shown to consist almost solely of b-sheet structure and was found to be highly resistant to proteolytic, chemical, acid and thermal denaturation.8,12 Mature PQC is a monomeric glycoprotein of 266 amino acid residues with a molecular mass of about 33 kDa.4,12 Furthermore, this enzyme exhibits catalytic activity over the broad pH range from 3.5 to 11.0.3,13 At neutral pH, unfolding of PQC was not observed either at elevated temperatures (up to 95 8C) or in the presence of high concentrations of chemical denaturants (7 M guanidium hydrochloride for instance).12 PQC is also quite resistant to proteolytic attack by various enzymes including bovine trypsin and bovine chymotrypsin as well as the four papaya cysteine proteinases.8 Based on present knowledge, human QC may possibly share the scaffold of the bacterial aminopeptidases, although all the putative QCs identified so far in bacteria are homologous to PQC. Human QC contains one zinc ion per molecule as analyzed by atomic absorption spectroscopy.14 Several heterocyclic compounds (imidazole, triazole and tetrazole rings) were recently identified as competitive inhibitors of human QC, due to the presence of nitrogen atoms that are probably good candidates to coordinate to the active-site zinc ion of human enzyme.15,16 The fact that human QC could be inhibited and inactivated by heterocylic chelators suggests that the presence of the zinc ion is essential for catalytic competence in the mammalian enzyme.17 In marked contrast, PQC is not inhibited at all by heterocylic chelators.12 Thus, altogether, mammalian and plant QCs are generally considered to belong to two distinct structural enzyme families.4,10 Herein they will be classified for clarity as type 1 QCs (plant, bacterial and parasite enzymes) and type 2 QCs (mammalian enzymes). While mammalian QCs are implicated in the maturation of numerous neuropeptides and

cytokins, such as thyrotropin-releasing hormone (TRH) and gonadotropin-releasing hormone (GnRH),18–20 the physiological function of plant QCs remains poorly characterized.21 Recently, it has been suggested that PQC could be involved in plant defense mechanisms,22 an hypothesis supported by the observation that the quantity of PQC expressed in the plant laticifers strongly increases as a result of repeated mechanical woundings.23 The role of type 1 QCs in both bacteria and parasites is unknown. ˚ We report here the crystal structure at 1.7 A resolution of the C. papaya glutaminyl cyclase enzyme. This first QC three-dimensional structure reveals that the papaya enzyme folds as a fivebladed b-propeller with the central tunnel filled by a zinc atom. This structure is thought to be representative of that of type 1 QCs from plants, bacteria and parasites.

Results Structure determination To solve the crystallographic structure of PQC, we relied on the availability of the pure enzyme extracted from the latex of C. papaya,12 since it is known that expression of this glycoprotein in a recombinant form gives only very poor yields.4 The crystallographic structure of PQC was solved using the single-wavelength anomalous diffraction (SAD) phasing technique with sulfur and zinc as the anomalous scatterers (266 residues, four sulfur atoms and one zinc ion leading to a calculated Bijvoet ratio of about 1%, see Materials and Methods). We turned towards this method after encountering difficulties in determining the structure by the multiple isomorphous replacement method. Since we expected initially only sulfur as the anomalous scatterer, we first tried to maximise the anomalous signal of sulfur atoms and thereby diffraction data were collected at the ˚ (7 keV) wavelength. The heavy-atom sites 1.77 A were located with the ShelxD program24 and the result suggested the presence of a stronger anomalous scatterer identified as zinc by subsequent atomic absorption and X-ray fluorescence studies (data not shown). Atom types were corrected accordingly, sites were refined and initial phases were calculated and further

459

Papaya Glutaminyl Cyclase Crystal Structure

improved by density modification with Sharp.25 The identified anomalous scatterer sites clustered in two groups related by a rotational noncrystallographic symmetry, consistent with the presence of two molecules in the crystal asymmetric unit, with a solvent content of 47% ˚ 3/Da). (Matthews coefficient of 2.1 A In order to measure higher resolution data, a data set was subsequently collected at the ˚ wavelength. The complete model was 0.95 A ˚ resolution. built stepwise and refined to 1.7 A Data collection and crystallographic refinement statistics are shown in Tables 1 and 2, respectively. Overall structure and topology Most of the protein backbone and side-chains are well defined. The final model comprises 250 residues (Arg5-Leu254) for the two molecules in the asymmetric unit. PQC folds into a five-bladed b-propeller (Figure 1). It exhibits a cylindrical ˚ shape characterized by a diameter of about 40 A ˚ and a height of 25 A. The five twisted b-sheets are arranged radially around a central tunnel for which the molecular environments at both extremities are very different. While the tunnel on the N-terminal face is widely open and surrounded by several long loops, its C-terminal entry is very narrow and filled with buffer molecules (Figure 1). A tris(hydroxymethyl) aminomethane (Tris) molecule is located at that position in monomer A while an acetate ion is observed in monomer B. Interestingly, in the known propeller structures the conical shape of the tunnel is often reversed, i.e. the entry located on the N-terminal face of the tunnel is narrower than the C-terminal side.26 In these structures the narrow face of the tunnel often bears the active site. The two monomers of PQC pack in the asymmetric unit in such a way that the C-terminal

Table 1. Data collection and processing statistics Energy ˚) Wavelength (A Anomalous data ˚) Resolution (A Space group Cell dimensions ˚) a (A ˚) b (A ˚) c (A Mosaicity No. unique reflections Completeness (%) I/s(I) Redundancy Rsymb a

7 keV 1.77 Yes 50–2.05 (2.17–2.05)a P212121

13 keV 0.95 No 50–1.7 (1.8–1.7) P212121

63.84 82.79 110.18 0.358 70,961 (10,987) 99.2 (96.9) 22.8 (9.7) 7.2 (6.7) 6.0 (14.5)

62.82 81.20 108.17 0.224 61,499 (9594) 99.9 (100) 41.6 (15.3) 28.9 (29.3) 6.0 (27.9)

Values in parentheses pertain to the outermost shell of data. Rsym Z SjIKhIij=SI, where I is the observed intensity, hIi is the statistically weighted average intensity of multiple observations of symmetry-related reflections. b

Table 2. Model statistics ˚) Resolution limits (A No. of protein atoms No. of sugar atoms No. of ligand atoms No. of water oxygen atoms No. of reflections in refinement No. of reflections in test set R-factor (%)a Rfree (%)b r.m.s.d. bond angles (8) ˚) r.m.s.d. bond lengths (A ˚) r.m.s.d. on Ca between monomers (A ˚ 2) B factors for monomers A–B (A For main-chain atoms For all atoms Ramachandran plot for monomers A–B Most favored (%) Additionally allowed (%) Generously allowed Disallowed

50.0–1.70 4126 88 14 896 58,424 3075 16.6 20.8 1.937 0.02 0.214 16.6–21.1 17.9–22.3 190 (85.6)–185 (83.3) 30 (13.5)–35 (15.8) 1: E69–1: E69 1: L40–1: L40

a R-factorZ SjjFo jKjFc jj=SjFo j, where jFoj and jFcj are the observed and calculated structure factor amplitudes, respectively. b Rfree was calculated with 5% of the data.

side entry of the central tunnel in monomer B is hidden by the other monomer and thereby completely inaccessible to the solvent (Figure 1). This molecular packing explains the differences in ligand binding at the central tunnel C-terminal cavity between monomers A and B. The interface between the monomers, with an area of about ˚ 2, is somewhat fuzzy with no significant 1000 A specific side-chain-driven contact. The complexed acetate ion is hydrogen bonded to residues Asn155 and Lys225 of monomer B and to Lys107, which is contributed by monomer A. Crystallization attempts omitting sodium acetate from the crystallization buffer were completely unsuccessful and led to protein precipitation, underlying the role of the acetate ion in the packing of molecules within the crystal. Sedimentation and chromatographic studies have indicated that PQC exists in a monomeric form.3 The binary complex found here in the crystallographic unit is thus presumed to result from the crystallization process and does not reflect the physical state of the enzyme in solution. Plant proteins are commonly N-glycosylated on asparagine residues that belong to a consensus sequence motif Asn-Xaa-Ser/Thr, where Xaa is any amino acid other than proline or aspartate.27 In the PQC primary sequence, this motif is only found at N101 and as expected, the X-ray structure shows only one N-glycosylation on N101. The electron density observed in this region allowed us to trace the first residues of the glycan chain in both monomers. We assigned it as (AsnND2-)GlcNac-GlcNac-mannose, with a fucose unit bound via an a1-3 link to the first GlcNac unit. The mannose unit has been built only in monomer A.

460

Papaya Glutaminyl Cyclase Crystal Structure

Figure 1. Ribbon drawing of the PQC structure. (a) Illustration of the crystal dimer and (b) stereo view of the PQC molecule, viewed along its pseudo 5-fold axis. Tris, acetate and sugar molecules are shown in ball-and-sticks and labelled. Zinc atoms are represented in orange. The disulfide bond is depicted in yellow. N and C termini are indicated.

The PQC structure: a modular architecture A typical b-propeller scaffold consists of four antiparallel stranded b-sheets, designed as blades, repeated four to eight times around a central axis.28,29 The polypeptide chain passes successively through each blade and progresses into it outwards from the central axis, from strand b1 to strand b4 (Figure 2). The crystal structure of PQC consists of five blades. As a consequence, in each blade, all loops between strands b1 and b2, and between strands b3 and b4 are positioned on the same face of the propeller, surrounding the wide-mouthed N-terminal entry side of the central tunnel.

Inversely, all loops connecting strands b2 to b3, as well as the loops connecting two adjacent blades between strand b4 and strand b1 of the next blade are found on the other face, around the narrow C-terminal side exit of the tunnel (Figure 2). Moreover, a b-hairpin (residues 145–151) and two a-helices (residues 190–200 and 246–250) are identified in PQC as additional structures. These additional secondary structures are positioned close to the polypeptidic N and C termini (Figure 1) and contribute to the polypeptide chain closure within the PQC b-propeller fold. Predictions of a b-propeller fold can generally be inferred from the presence of internal repeated

461

Papaya Glutaminyl Cyclase Crystal Structure

C 252L

55K

51L

32D

30E

97S

93R

76E

251C

56V

50A

33T

29A

98N

92D

77K

74L

57E

49V

34L

28Y

99I

91Y

78L

73L

58N

48Q

35F

27V

100K

90I

79Y

72T

247I

59I

47R

36E

26L

101N

89F

80Q

71L

256Y

60H

46V

37S

25G

102F

88G

81V

70G

245G

61K

45S

38T

24Q

103T

87I

82V

69E

62M

44S

39G

23T

104H

86N

83W

68G

250H 249R 162G

179A

183T

160I

163E

178S

184L

159Y

164V

177I

185L

158E

165W

176R

186G

5R

157L

166A

175A

187W

6V

156E

167N

174I

188I

7Y

237L

155N

168I

173C

189L

8I

236H

154L

169W

172D

Blade 4

N

238V

3-joint molecular velcro

9V

235L

191P

10E

234K

219R

214D

192N

11V

233I

220I

213W

193L

12L

232E

221F

212A

190L 153R

248E

194R

13N

231F

222V

230L

223T

210G

197L

15F

229K

224G

209N

198I

16P

228P

225K

208L

199D 200E

207V

17H

Blade 5

Blade 2 137K

149H

146Y

138L

132D

118I

115D

150R

145K

139I

131I

119L

114T

140K

130E

120Y

113A

141K

129Y

121G

112L

142H

128L

122S

111G

143N

127I

123D

144V

126S

151V 14E

196K

Blade 1

211I

195K

67F

22F

Blade 3

Figure 2. Topology diagram of PQC structure computed by PROMOTIF.64 The PQC sequence number and one-letter code are shown within circles and rectangles, for a-helix and b-strand, respectively. Hydrogen bonds are represented by brown arrows. The three-joint molecular velcro is illustrated by way of three coloured circles. The five blades are labelled.

sequence motifs.26 Although its amino acid sequence does not exhibit any such repeats, a propeller fold has also been predicted for PQC9 on the basis of the experimentally determined secondary structure content8 and of the high rigidity and stability of a propeller,28 two biophysical characteristics of the PQC protein. An original molecular velcro The circular structure of b-propellers is generally closed by the association, into the last blade, of b-strands originating both from the N and C-terminal ends.26,30 This so-called “velcro” closure is achieved by the insertion of the first N-terminal b-strand into the last C-terminal b-sheet. However in the PQC structure the classical molecular velcro is further strengthened resulting in an original super-molecular velcro with three levels of jointure (illustrated on Figure 2), which undoubtedly provides additional stability to the structure. The antiparallel hydrogen bonding pattern between the N and, C-terminal strands is lengthened to 11 residues and, as a consequence, a part of the first N-terminal b-strand is also included in a parallel hydrogen bonding to a b-strand of blade 4 (Figure 2). The last two blades in the PQC structure are thereby linked, producing a large eightstranded b-sheet with a topology of (K4,C1,C 1,C1,C4, K1, K1) according to Richardson’s nomenclature.31 The existence of this super molecular velcro is made possible by the presence of several b-bulge local irregularities in the b sheet pattern of hydrogen bonding (Figure 2), resulting in additional twist in the b-strand structures. In this respect, the six-bladed propeller structure of phytase, a thermostable enzyme, also appears to have a non-canonical molecular velcro, which consists, in fact, of two velcro systems.32 PQC contains two cysteinyl residues located at positions 173 and 251 in the amino acid sequence. The

structure shows that they form a disulfide bridge wherein the two Sg atoms are separated by a ˚ . This fully buried disulfide distance of 2.08 A bond obviously provides additional structural stability to the circular array of the enzyme, since it prevents the last loop (Arg239–Asp244) and the last a helix from being flexible and potentially susceptible to be degraded by the papaya proteinases. In a previous study,12 the presence of two moles SH per mole of enzyme was described, suggesting the absence of a disulfide bond, an observation that is in disagreement with our crystal structure. In this study, the PQC was purified from a mixture of cysteine proteinases that were blocked as S-methylthiolderivatives. Zerhouni et al.12 expected thus to have a PQC-S-methylthiolderivative enzyme and thereby before thiol determination have incubated the enzyme in DTT (2.5 mM) during 30 min at room temperature. Disulfide bonds may have been reduced by this procedure and then quantified as free thiol by mistake. Our structural result clearly reveals the presence of a disulfide bond in PQC. PQC is a metalloprotein While interpreting the results of ShelxD in order to locate the positions of the anomalous scatterers in PQC we observed the presence in the asymmetric unit of two strong anomalous scatterers that could not be assigned to sulfur atoms. Two different spectroscopic methods, namely atomic absorption and X-ray fluorescence, were used to identify these unexpected scatterers as zinc ions. Both methods also indicated the weaker binding of copper ions. Neither zinc nor copper was added in the protein preparation or crystallization buffer. Zinc was included in the structure, as the predominant metal identified, although the presence of traces of copper raises the possibility that copper could

462

Figure 3. Illustration of the zinc-binding site, represented in the 2FoKFc electron density map contoured at the 1.0s level. The zinc ion is shown as an orange sphere and the water molecule as a pink sphere. Residues around the zinc ion (Leu26, Glu156, Leu157, Glu158, Ile211) are depicted and labelled.

Papaya Glutaminyl Cyclase Crystal Structure

and both solutions equally populated. In addition, both entries possess a negative electrostatic surface potential, illustrated in Figure 4, even if the N-terminal side is more negatively charged. In the PQC crystal structure, a Tris molecule or an acetate ion is observed in the pocket of the central tunnel located at the C-terminal side. It was previously reported that Tris had a slightly inhibiting effect on PQC,3,16 but without further characterization. In order to gain insights into the effect of Tris on PQC, kinetic analyses of the inhibition of glutamine cyclization by Tris were performed, showing that Tris is a competitive inhibitor of PQC, with an inhibition constant (Ki) of about 29 mM. Thus, the Tris binding site identified in the PQC crystal structure very likely represents the active site of PQC. This result is indeed strongly supported by a multiple sequence alignment of type I QC sequences (Figure 5). Residues that belong to the N-terminal side entry, i.e. four glutamate residues (30, 158, 215 and 216), two lysine residues (117 and 218), Gln215, Asn75 and Asp115 are not conserved between the type 1 QCs, in sharp contrast with

substitute, to some extent, for zinc in the PQC structure. The zinc coordination number in PQC is six, and its binding site therefore adopts an octahedral coordination geometry32 with, as in-plane coordinating atoms, the three backbone carbonyl oxygen atoms of Leu26, Leu157 and Ile211 and one water molecule and, as out-of-plane coordinating atoms, the carboxylate groups of Glu156 and Glu158 (Figure 3). The carboxylate group of Glu156 makes a bidentate binding with its two oxygen atoms simultaneously coordinated to the zinc atom. The observed distances between the zinc atom and the ˚ (Leu26 O), 2.27 A ˚ coordinating atoms are 2.18 A ˚ (Ile211 O), 2.29 A ˚ (Wat O), 2.22 A ˚ (Leu157 O), 2.24 A ˚ ˚ (Glu158 OE1) and 2.31 A and 2.46 A (Glu156 OE1 and OE2). Identification and description of the putative catalytic site The active site of most b-propeller enzymes is located at one entrance of their central tunnel,33 either at their N-terminal or C-terminal side according to the case. As a first approach to locate the active site of PQC, we have performed prediction of putative active sites with the program PASS34 as well as molecular docking studies using the AutoDock program.35 Two unique low-energy solutions have emerged from these analyses. One of them places the active site at the N-terminal entry of the central tunnel and the other one at the C-terminal entry. The virtual docking did not allow us to distinguish between the two putative sites, since the docking energies were comparable

Figure 4. Surface representation of (a) the N-terminal and (b) C-terminal entry of the PQC b-propeller structure. The molecular surface is coloured according to the electrostatic potential computed with DELPHI and displayed with GRASP.65 Colour codes for the electrostatic potentials are: K3kT/e (red), 0kT/e (white), and C3kT/e (blue).

Papaya Glutaminyl Cyclase Crystal Structure

463

Figure 5. Multiple sequence alignment of 17 type 1 QC proteins. The QC proteins are labelled by their accession code. These are O81226, Carica papaya; Q84WV9, Arabidopsis thaliana; Q5VRH1, Oryza sativa; Q9A519, Caulobacter crescentus; Q5NLA9, Zymomonas mobilis; Q7X2U3,Acidobacteria bacterium; Q8ZAD5, Yersinia pestis; Q66G23, Yersinia pseudotuberculosis; Q8P8M4, Xanthomonas campestris; Q8PK56, Xanthomonas axonopodis; Q6MA45, Parachlamydia sp.; Q9RY39, Deinococcus radiodurans; Q5Z254, Nocardia farcinica; Q8FR67, Corynebacterium efficiens; Q8NS57, Corynebacterium glutamicum; Q9A125, Streptococcus pyogenes; Q6NIJ7, Corynebacterium diphtheriae. The secondary structure elements according to the PQC structure are indicated below the alignment by coloured boxes, in red for a-helices and in green for b-strands. The first lines contain the PQC sequence numbering. Light yellow background positions are highly conserved positions, which often correspond to packing hydrophobic residues; orange background are residues that coordinate the zinc atom; the glycosylated asparagine is in violet background colour and cysteine residues forming the disulfide bond are in blue background. Residues of the putative active site are in dark yellow background. Note that two PQC sequences of parasites, Plasmodium falciparum (Q8IL03) and Plasmodium yoelii yoelii (Q7R9G6), were omitted from this alignment because of the presence of large gaps in their sequence alignment.

residues found in the C-terminal site entry (see Figure 5). The C-terminal side entry is lined up by the following nine highly conserved residues (Figure 5): Trp83, Trp110, Trp169, Gln24, Glu69, Lys225, Asn155, Phe22 and Phe67. The three tryptophan residues form the bottom of an hydrophobic pocket in which substrate molecules could be embedded, in a way similar to the Tris molecule (Figure 6). The three indole groups are arranged in to form three faces of a rectangular box with the planes of Trp83 and Trp169 being almost parallel with each other and that of Trp110 is perpendicular to both. Such a disposition, which is also observed in the case of the betaine binding site of the protein ProX from Escherichia coli,36 could possibly create a favourable environment in order to facilitate the intramolecular cyclization of the glutaminyl residue.13 As a first approach to gain insights in the glutamine cyclization process, we have tentatively fitted glutamine in a “transition state-like” conformation based on the observed contacts between the QC and the Tris molecule. Although speculative in the absence of any co-crystallization with a substrate analogue, this approach allows us to pinpoint residues potentially involved in the

catalytic process, and from their geometrical disposition, to speculate about their respective role in catalysis. Four highly conserved residues, namely E69, Q24, N155 and K225, appear crucial for Tris binding. The amino group N1 of the Tris molecule (see Figure 6(b) for atom nomenclature) is hydrogen bonded to E69OE1 and Q24OE1. The Tris hydroxyl O2 is hydrogen bonded to Q24NE2, N155OD1 and E69OE1, while the hydroxyl O1 is hydrogen bonded to N155 ND2, Q24 NE2 and K225 NZ. A glutamine residue has been positioned on the basis of the positions of the Tris amino group for its a-amino group and of the positions of the Tris hydroxyl groups O1 and O2 for its side-chain groups OE1 and NE2, respectively. In such modelling, E69 OE1 and OE2 interact with the side-chain NE2 and the a amino group, respectively. The NE2 and OE1 side-chain groups are hydrogen bonded to N155 OD1 and ND2, while the side-chain OE1 also interacts with K225 NZ. Q24 can also contribute to hydrogen bonding with the substrate in a way similar to that observed for the Tris molecule. The polypeptidic chain of the substrate would extend in the direction of W110, W169 and W83, and this direction indeed points outwards to a cavity filled with well-defined water molecules. This model

464

Papaya Glutaminyl Cyclase Crystal Structure

Figure 6. Stereo view of the observed interactions between the Tris molecule and PQC (monomer A). The 2FoKFc electron density map for the Tris molecule is contoured at the 1.5s level and is shown in cyan. (b) Schematic representation of the observed interactions between Tris and PQC. The two-dimensional representation for the interaction between the Tris molecule and PQC was obtained using Ligplot67 where atoms participating in hydrogen bonds (with distances) and hydrophobic contacts are indicated.

building is consistent with the proposed catalytic mechanism,13 where K225 might help the nucleophilic attack on the side-chain carbonyl group by the nitrogen atom of the a-amino group of the substrate, thus forming the charged transition state. Protonation of the leaving group can occur as a result of the enzyme-assisted transfer of the proton from the a-amino group, with the help of the acidic group E69. Q24 and N155 could contribute to hydrogen bonds in order to stabilize the transition state during glutamine cyclization (Figure 7).

Discussion We have determined the crystal structure at a ˚ of the C. papaya glutaminyl resolution of 1.7 A cyclase enzyme. The structure was solved by the SAD method on the basis of the anomalous signal of sulfur and zinc atoms. The PQC structure represents another example of structures determined using this technique, which illustrates that SAD

phasing of a native protein using sulfur and/or naturally bound metals may become a method of choice for obtaining initial phases. PQC folds into a five-bladed b propeller, an architecture that is common to all type I QCs. Type I QCs include enzymes found in plants and bacteria, and in some parasites, such as Plasmodium, Parachlamydia and Yersinia. Type 2 QCs on the other hand are found in animals. The crystal structure of PQC further reveals that this plant enzyme is a metallo-protein, a so far unexpected feature. A large cavity, elongated along the pseudo 5-fold axis of the b-propeller fold, contains a zinc ion. Although buried in the structure and thus inaccessible to the substrate, this structural zinc contains one water molecule in its coordination shell. Such a feature is not unusual and has been observed in other structures containing a structural zinc ion.37 Finally, the enzyme also contains a disulfide bridge. Amongst the different b-propeller structures so far reported, those containing five blades are the most recently revealed. The first known five-bladed

Papaya Glutaminyl Cyclase Crystal Structure

Figure 7. Schematic representation of (a) the observed interactions between Tris and PQC and (b) putative interactions between a glutamine residue in a “transition state-like conformation” and PQC.

structure dates from 1999 and was found in tachylectin-2, a protein belonging to the lectin family and involved in innate immunity.35 Thereafter, a-L-arabinanase 43A38 and apyrase, a human platelet function inhibitor,39 were identified as fivebladed propeller proteins. To our knowledge PQC represents thus the fourth known five-bladed bpropeller. While the central tunnel of tachylectin-2 is filled with water molecules arranged in a closed pentagonal cage conformation,40 apyrase, and probably arabinanase40–42 contain calcium as a central ion. These calcium ions are located in the central cavity axis of the b-propeller fold. Since the calcium ions do not participate to the biological functions, it is assumed that their role is confined to help stabilization of the central tunnel by tethering the five-bladed b-propeller. The observation that the thermostability of human apyrase decreases upon removal of calcium ions by EDTA chelation,39 argues in favour of this hypothesis. PQC is known to be highly resistant to proteolytic, chemical, acid and thermal denaturation.8,12 It exhibits catalytic activity over the broad pH range 3–11.3,13 At neutral pH, unfolding of PQC is not observed either at elevated temperatures (up to 95 8C) or in the presence of high concentrations of chemical denaturants (e.g. 7 M guanidium hydrochloride).12 Furthermore, PQC is quite resistant to proteolytic attack by enzymes such as bovine

465 trypsin, bovine chymotrypsin as well as to the four papaya cysteine endopeptidases.8 By analogy to the situation of apyrase and arabinanase, the zinc ion of PQC may well play a role in stabilizing the enzyme conformation. The overall fold also contributes in a general way to the protein stability, since both extremities of the polypeptide chain are encompassed into the global organization of the propeller (Figure 2), in such a way that the protein cannot be simply unwrapped from one end. The original molecular velcro in which the last two blades in the PQC structure are thereby linked, producing a large eight-stranded b-sheet, may also contribute to the protein stability. In the case of arabinanase, which also possesses an original closure mode, N-terminal residues are involved in hydrogen bonds and hydrophobic interactions implying the first and the last blades. It has been demonstrated that these interactions considerably contribute to the protein stability.41 Our structural analysis reveals that the high stability of the PQC can also result from other structural features. First, many tightly packed hydrophobic contacts established between the blades may also contribute to the PQC stability. Such contacts implicate the side-chains of residues of leucine 12, 26, 34, 71, 74, 78, 96, 112, 119, 128, 154, 157, 184, 208, 230, of isoleucine 90, 131, 160, 174, 177, 211, of phenylalanine 15, 22, 35, 67, 89, 102, 136, 221 and of valine 27, 46, 81, 82, 144, 164, 207, 222. Most of these residues, highly conserved among type I QC sequences (see Figure 5) are thus anticipated to form crucial interactions required for the b-propeller fold of these enzymes. The PQC structure is also stabilized through extensive hydrogen bond networks, with 188 backbone hydrogen bonds (accounting 75.2 hydrogen bonds per 100 residues) and 146 residues (58.2% out of the total) participating in regular secondary structures. Finally, with the exception of the loops at the N-terminal end of both a helices, the PQC structure contains only short loops, which may explain thereby the high resistance that the protein exhibits towards proteolysis, as previously demonstrated.8,12 Type 1 and type 2 QCs share similar catalytic properties, as revealed by the substrate specificity pattern and pH dependence of catalysis.16 In addition, both QC types possess one zinc ion per protein molecule. However, in human QCs it has been demonstrated that the zinc atom is involved in its enzymatic function,14,16 suggesting divergent enzymatic mechanisms between both types of enzymes. Crystal structures of human QC have been reported recently.43 As expected,9 human QC shares the scaffold of bacterial aminopeptidases, an a/b open-sandwich topology with a central b-sheet surrounded by several a-helices.43 Interestingly, the active site, which is created by six loops, is lined by several aromatic residues, which may help in substrate binding, as supported by site-directed mutagenesis studies.43 However, these aromatic residues are not arranged into an aromatic box geometry as found in the PQC active site. Further-

466 more, comparison between our speculated catalytic mechanism for PQC and that of human QC (Figure 5 of Huang et al.43) shows some striking similarities. The role of E69 of PQC should be played by E201 in human QC, and the action of zinc catalytic atom of human QC could be assumed by K225 in PQC with the participation of Q24 and N155 side-chains. In human QC, the zinc atom is far from the centre and lies on the bottom of the active site pocket, which is mainly created by the six loops. Studies on the reaction mechanism of PQC have shown that catalysis does not proceed via an acylenzyme mechanism, but rather favour a simple intramolecular cyclization of the glutamine residue to the pyroglutamic acid residue.13 From a structural view the glutamine cyclization reaction consists of the formation of a five-membered ring containing an internal peptide bond. The mechanism proposed consists of the following main steps:13 an initial intramolecular nucleophilic attack on the side-chain gC]O carbon by the nitrogen of the a-amino group, followed by the transfer of a proton from the a-amino group to the nitrogen atom of the amide group, facilitated by an acidic group present in the enzyme active site, and finally the expulsion of the ammonia-leaving group promoted by the same or another acidic enzyme group. Our preliminary analysis of the putative active site of PQC, identified by the presence in the crystal structure of a competitive inhibitor Tris molecule, combined with sequence analyses, pinpointed residues potentially involved in the catalytic mechanism. Although speculative, we tentatively modelled a glutamine residue into the active site, in order to gain a first insight into the catalytic process of PQC. In such modelling, the initial intramolecular nucleophilic attack would be facilitated by K225 as a result of its interaction with and polarizations of the gC]O substrate group. In human QC, the role of K225 should be played by the catalytic zinc atom.43 The proton transfer could be assisted by E69, the only acidic group present in the PQC active site. E69, corresponding to E201 in human QC,43 could also facilitate the expulsion of the ammonia-leaving group. In the proposed catalytic mechanism of PQC, the side-chain carbonyl and a-amino groups must be at different sides of the plane defined by the a, b, and g carbon atoms.13 The positions of K225 and E69 in the structure of PQC are in agreement with such a model. The transition state would be further stabilized via additional interactions with N155 and Q24, while the active site could be delimited by the three tryptophan residues 169, 110 and 83. The physiological function of PQC remains poorly characterized.21 A recent study has shown that regularly wounded papaya fruits accumulate, in their latex, large amounts of PQC, while the enzyme is not present in the latex of fruits that are tapped for the first time. The observation that PQC belongs to the wound-induced protein family that contains plant defense enzymes remains unexplained, although ammonia released by the PQC

Papaya Glutaminyl Cyclase Crystal Structure

activity may have a negative effect on bacteria and thus play a defensive role.22 Also, the possibility that PQC might bear another catalytic activity, still to be demonstrated, cannot be excluded. In humans, the QC activity can be associated to pathological processes such as osteoporosis and amyloidotic diseases. Several nucleotide variations in the region of pituitary QC gene indeed are strikingly correlated with osteoporosis susceptibility in adult women.44 Also it is well stated that among all prominent amyloid b peptides that contribute to Alzheimer ’s disease, those that contain pyroglutamic acid at the N-terminus are the most abundant ones.45,46 These N-terminal pyroglutamic acids probably result from the enzymatic cyclization of glutamyl residues, since the gene sequence of amyloid b fragments indicates that glutamic acid is the N-terminal residue and since the cyclization of this amino acid does not spontaneously occur either in vitro or in vivo.47 Interestingly, human QC and PQC were recently described in vitro as able to catalyze also N-terminal glutamate cyclization, but with a relatively inefficient rate as compared to Gln conversion.47 Even if the implication of human QC in Alzheimer pathology remains a hypothesis, all studies to understand in general the function of QC, of both type 1 and type 2, could be important in this context.

Materials and Methods X-ray diffraction data Protein purification and crystallization procedures were as reported.48 Two sets of diffraction data were collected on a single crystal using the beamline BM14 at the European Synchrotron Radiation Facility (Grenoble, France) equipped with an MAR CCD225 detector: a high ˚ and a data set at a lZ energy set (13 keV) at a lZ0.9537 A ˚ (7 keV) to maximize the sulfur anomalous signal 1.77 A and with the constraint to preserve a correct signal intensity. The low-energy data set consisted of two passes of 3608 rotation range with two different k angles in order to maximize data redundancy. All data were integrated, scaled and merged with the XDS suite of programs.49 The relevant data collection parameters are listed in Table 1. Phase determination ˚ wavelength were used to Data collected at the 1.77 A locate the heavy-atom sites. Two strong anomalous scatterers and eight weaker ones were detected by using the program ShelxD.24 Consistent with two molecules in the asymmetric unit, these latter sites correspond to the four sulfur atoms from the two methionine residues and the two cysteine residues. Both atomic absorption and X-ray fluorescence spectroscopies identified the two strongest anomalous scatterers as zinc ions. According to Hendrickson & Teeter’s formula,50 the expected Bijvoet average ratio is given by hjDFGji=hFiZ ð2NA NP Þ1=2 ðdfA00 =Zeff Þ. In the case of PQC, with NP of about 4400 atoms in the asymmetric unit, with the average ZeffZ6.7 electrons and assuming NAZ8

467

Papaya Glutaminyl Cyclase Crystal Structure

sulfur atoms with dfA00 Z 0:72 electron at a wavelength of ˚ , this ratio is 0.65%, a value close to the limit of lZ1.77 A application of the S-SAD phasing method.51 However, considering zinc ion scatterers (dfA00 Z 0:87 electron at lZ ˚ ), the estimation of the Bijvoet ratio hjDFGji=hFi, 1.77 A increases to about 1%. The autoSharp program suite25 was used to refine the heavy-atom sites and to derive the protein initial phases ˚ wavelength data set using the SAD from the 1.77 A 52 method. Following solvent flattening and density modification, the initial electron density map was subjected to automatic model building with wARP53 ˚ wavelength data set. using the 0.95 A Model building and refinement The refinement procedure was carried out initially using the program CNS 54 and involved conjugate gradient minimization procedures with maximum likelihood. Water molecules were added with the standard water_pick script in CNS. Coordinates and stereochemical parameters of sugar moieties were downloaded from the Hetero-compound Information Center (HIC-UP, Uppsala, Sweden)55 and the carbohydrate chains were built with the TURBO-FRODO molecular modelling program,56 based on the work published about structures of N-glycans in C. papaya.57,58 The progress of refinement and rebuilding was monitored with the free R-factor (Rfree) computed from a randomly omitted 5% of the observed reflections. The refinement with CNS proceeded using automatically determined X-ray/ geometric (Wa) and X-ray/Brestraints (Rweight) weights and ended with values for R and Rfree of 20.8 and 24.1%, respectively. An acceptable but rather poor geometry was observed for the models, with only 81.4% of the residues in the most favored region of the Ramachandran plot. The refinement was pursued by optimizing the Wa and Rweight terms with the optimize scripts provided in CNS. Refinement using the optimized terms allowed us to lower the R and Rfree factors to 19.9% and 23.7%, respectively, with a slight improvement in model geometry, as illustrated by 83.3% of the residues found in the most favored region of the Ramachandran plot. These values were not estimated as outstanding, regarding the high quality of the diffraction data. In order to complete the refinement, we turned to the refinement software Refmac559 from the CCP4 suite.60 The set of test reflections was kept identical with that used in CNS. Parameters for ligands and carbohydrates were taken from the dictionaries provided with Refmac5. 61 Restrained refinement with Refmac5 resulted in an improved electron density map that allowed us to correct some side-chain conformations, to position the O3 oxygen of the Tris ligand, to correct the conformation of the fucose rings found in the two glycosylated chains, and to add a mannose unit to the glycosylation site in monomer A. The final refinement and model statistics are provided in Table 2. The four first N-terminal residues and the last 12 C-terminal residues were not visible in the electron density map and are not included in the final model. Final R and Rfree are 16.6% and 20.7%, respectively, with 84.5% of the residues in the most favored region of the Ramachandran plot, and 248 residues out of 250 found in each monomer are in the most favored or additional allowed regions of the plot. One residue, Glu69, lies in the generously allowed region, just outside the additional allowed region, in each monomer. Glu69 is involved in either Tris (monomer A) or acetate (monomer B) binding. Also, one residue, Leu40, lies in the disallowed region of

the plot, in each monomer. Its conformation has been carefully checked and confirmed by the very well defined electron density at that position both in monomers A and B. The r.m.s. deviation between the main-chain atoms of ˚ . The the two monomers in the asymmetric unit is 0.218 A ˚, average displacement for main-chain atoms is 0.186 A ˚ , a difference and the largest displacement is 0.646 A observed for Leu84, which in monomer B participates in the monomers packing within the asymmetric unit. Structure analysis The structure was checked and analyzed with the PROCHECK,62 DSSP,63 PROMOTIF64 and GRASP65 programs. In order to identify the active site, prediction of binding pockets was performed with the PASS program34 and docking screening was carried out with AutoDock3.0.5,35 using the graphical interface AutoDockTools. For this purpose, all water and solvent molecules were removed from the crystal structure and polar hydrogen atoms were added. Kollman charges66 were assigned to protein atoms. Mass-centered energy grip maps were generated by the AutoGrid3 program35 ˚ spacing and for the whole protein (dimenwith 0.375 A sions of 120!120!120 point). The Lamarckian genetic algorithm methods were applied for minimization using default parameters. As recommended for blind docking,67 150 runs were performed with 10 million generations, 10 million energy evaluations and a population size of 250. Several molecules were used as ligand: a Tris molecule, a glutamine amino acid, a proline amino acid and a 5-oxo-proline moiety. Figures 1, 3 and 6 were drawn using a combination of MOLSCRIPT,68 BobScript69 and Raster3D70 programs. Figure 2 was obtained with PROMOTIF64 and Figure 4 with GRASP66 using DELPHI for the electrostatic potential calculations. Figure 7(a) was generated using the Ligplot program.71 Kinetic experiments PQC activities were assayed spectrophotometrically using the procedure described by Gololobov et al.13,15 Briefly, 5 ml of a PQC stock solution (1.4 mg/ml) was pipetted into a defined volume of 0.1 M phosphate buffer at pH 8.6. The reaction mixtures were made up in the spectrophotometer cuvettes and kept in the thermostated jacket of the instrument (25 8C) for at least 15 min. The reaction was initiated by the addition of 1–100 ml of the L-glutamine tert-butyl ester (GlntBE, Sigma) stock solution (50 mM). The substrate was prepared in water and pH was adjusted to 8.4–8.8. In all experiments, the final reaction volume was 1 ml. The release of pGlutBE was followed at 210 nm using a Perkin-Elmer lambda45 spectrophotometer. From the Eadie–Hofsteen plot (ratio of velocity and initial GlntBE concentration as a function of velocity), the Km value was determined to be 0.33 mM. In order to determine the type of inhibition and the Ki of Tris on PQC activity, a Dixon analysis was performed, consisting of a graph of the reciprocal of velocity against inhibitor concentration for different GlntBE concentrations. For this purpose, different volumes of a 1 M Tris solution stock were added in reaction mixtures 15 min before the addition of GlntBE substrate to initiate the reaction. The 1 M Tris solution was prepared in 0.1 M phosphate buffer and the pH adjusted to 8.6. The Tris concentrations were 1, 5, 10, 15, 20, 25 and 30 mM, whereas three different GlntBE concentrations were used,

468 0.25 mM, 0.4 mM and 0.5 mM. The three lines (one for each substrate concentration) intersected each other above the X axis, which revealed a competitive (or mixed) inhibitor pattern. The Ki value was calculated to be 0.29 mM. Protein Data Bank accession number The atomic coordinates and structure factors of PQC have been deposited in the RCSB Protein Data Bank with accession code 2FAW.

Acknowledgements We thank H. Drobecq for the mass spectrometry, T. Doneux for the atomic absorption measurements and Professor J.- P. Dehaye for help in kinetic analysis. C.B., B.C. are V.V. are at the Centre National de la Recherche Scientifique (CNRS), France. We thank the MRC-BM14 Staff and the EMBL staff for making the UK CRG BM14 beamline available to us. This work was supported, in part, by the CNRS and the Re´gion Nord-Pas de Calais through the CPER and FEDER. V.V. was a recipient of an Action The´matique et Incitative sur Programme Jeunes Chercheurs Grant from the CNRS. Y.L. and R.W. acknowledge the Communaute´ Franc¸aise de Belgique (Action de Recherche Concerte´e, Convention no. 02/07-289) for support.

References 1. Messer, M. (1963). Cyclization of L-glutamine and L-glutaminyl peptides. Nature, 197, 1299. 2. Messer, M. & Ottesen, M. (1964). Isolation and properties of glutamine cyclotransferase of dried papaya latex. Biochim. Biophys. Acta, 92, 409–411. 3. Messer, M. & Ottesen, M. (1965). Isolation and properties of glutamine cyclotransferase of dried papaya latex. C.R. Trav. Lab. Carlsberg, 35, 1–24. 4. Dahl, S. W., Slaughter, C., Lauritzen, R. C., Bateman, R. C., Jr, Connerton, I. & Pedersen, J. (2000). Carica papaya glutamine cyclotransferase belongs to a novel plant enzyme subfamily: cloning and characterization of the recombinant enzyme. Protein Expr. Purif. 20, 27–36. 5. Arri, K., Kobayashi, H., Kai, T. & Kokuba, Y. (1999). Degradation kinetics of L-glutamine in aqueous solution. Eur. J. Pharma. Sci. 9, 75–78. 6. Busby, W. H., Jr, Quackenbush, G. E., Humm, J., Youngblood, W. W. & Kizer, J. S. (1987). An enzyme(s) that converts glutaminyl-peptides into pyroglutaminyl-peptides. J. Biol. Chem. 262, 8532–8536. 7. Fisher, W. H. & Spiess, J. (1987). Identification of a mammalian glutaminyl cyclase converting glutaminyl into pyroglutaminyl peptides. Proc. Natl Acad. Sci. USA, 84, 3628–3632. 8. Oberg, K. A., Ruysschaert, J.-M., Azarkan, M., Smolders, N., Zerhouni, S., Wintjens, R. et al. (1998). Papaya glutamine cyclase, a plant enzyme highly resistant to proteolysis, adopts an all-beta conformation. Eur. J. Biochem. 258, 214–222.

Papaya Glutaminyl Cyclase Crystal Structure

9. Booth, R. E., Lowell, S. C., Misquitta, S. A. & Bateman, R. C., Jr (2004). Human glutaminyl cyclase and bacterial zinc aminopeptidase share a common fold and active site. BMC Biol. 2, 2–10. 10. Bateman, R. C., Jr, Temple, J. S., Misquitta, S. A. & Booth, R. E. (2001). Evidence for essential histidines in human pituitary glutaminyl cyclase. Biochemistry, 40, 11246–11250. 11. Schilling, S., Hoffmann, T., Rosche, F., Manhart, S., Wasternack, C. & Demuth, H.-U. (2002). Heterologous expression and characterization of human glutaminyl cyclase: evidence for a disulfide bond with importance for catalytic activity. Biochemistry, 41, 10849–10857. 12. Zerhouni, S., Amrani, A., Nijs, M., Smolders, N., Azarkan, M., Vincentelli, J. & Looze, Y. (1998). Purification and characterization of papaya glutamine cyclotransferase, a plant enzyme highly resistant to chemical, acid and thermal denaturation. Biochim. Biophys. Acta, 1387, 275–290. 13. Gololobov, M. Y., Song, I., Wang, W. & Bateman, R. C., Jr (1994). Steady-state kinetics of glutamine cyclotransferase. Arch. Biochem. Biophys. 209, 300–307. 14. Huan, K.-F., Liu, Y.-L. & Wang, A. H.-J. (2005). Cloning, expression, characterization of a glutaminyl cyclase from human bone marrow: a single zinc metalloenzyme. Protein Expr Purif. 2005 Sep;43(1):65– 72. Epub 2005 Mar 19. 15. Gololobov, M. Y., Wang, W. & Bateman, R. C., Jr (1996). Substrate and inhibitor specificity of glutamine cyclotransferase (QC). Biol. Chem. Hoppe-Seyler, 337, 395–398. 16. Schilling, S., Manhart, S., Hoffmann, T., Ludwig, H.-H., Wasternack, C. & Demuth, H.-U. (2003). Substrate specificity of glutaminyl cyclases from plants and animals. Biol. Chem. 384, 1583–1592. 17. Schilling, S., Niestroj, A. J., Rahfeld, J.-U., Hoffmann, T., Wermann, M., Zunkel, K. et al. (2003). Identification of human glutaminyl cyclase as a metalloenzyme. J. Biol. Chem. 278, 49773–49779. 18. Pohl, T., Zimmer, M., Mugele, K. & Spiess, J. (1991). Structure and functional expression of a glutaminyl cyclase. Proc. Natl Acad. Sci. USA, 88, 10059–10063. 19. Song, I., Chuang, C. Z. & Bateman, R. C., Jr (1994). Molecular cloning, sequence analysis and expression of human pituitary glutaminyl cyclase. J. Mol. Endocrinol. 13, 77–86. 20. Bockers, T. M., Kreutz, M. R. & Pohl, T. (1995). Glutaminyl-cyclase expression in the bovine/porcine hypothalamus and pituitary. J. Neuroendocrinol. 7, 445–453. 21. Amrani, A., Wintjens, R., Zerhouni, S., Azarkan, M., Nijs, M., Vincentelli, J. & Looze, Y. (1997). Conversion of the N-terminal glutaminyl residues of plant proteins into 5-oxoproline: a post-translation modification in quest of a physiological function. Recent Res. Dev. Phytochem. 1, 25–37. 22. El Moussaoui, A., Nijs, M., Paul, C., Wintjens, R., Vincentelli, J., Azarkan, M. & Looze, Y. (2001). Revisiting the enzymes stored in the laticifers of Carica papaya in the context of their possible participation in the plant defence mechanism. Cell. Mol. Life Sci. 58, 556–570. 23. Azarkan, M., Wintjens, R., Looze, Y. & BaeyensVolant, D. (2004). Detection of three wound-induced proteins in papaya latex. Phytochemistry, 65, 525–534.

469

Papaya Glutaminyl Cyclase Crystal Structure

24. Schneider, T. R. & Sheldrick, G. (2002). Substructure solution with SHELXD. Acta Crystallog. sect. D, 58, 1772–1779. 25. de la Fortelle, E. & Bricogne, G. (1997). Maximumlikelihood heavy-atom parameter refinement for multiple isomorphous replacement and multiwavelength anomalous diffraction methods. Methods Enzymol. 276, 472–494. 26. Paoli, M. (2001). Folds propelled by diversity. Prog. Biophys. Mol. Biol. 76, 103–130. 27. Kornfeld, R. & Kornfeld, S. (1985). Assembly of asparagine-linked oligosaccharides. Annu. Rev. Biochem. 54, 631–664. 28. Fu¨lo¨p, V. & Jones, D. T. (1999). b Propellers: structural rigidity and functional diversity. Curr. Opin. Struct. Biol. 9, 715–721. 29. Jawad, Z. & Paoli, M. (2002). Novel sequences propel familiar folds. Structure, 10, 447–454. 30. Neer, E. J. & Smith, T. F. (1996). G protein heterodimers: new structures propel new questions. Cell, 84, 175–178. 31. Richardson, J. S. (1981). The anatomy and taxonomy of protein structure. Advan. Protein Chem. 34, 167–339. 32. Ha, N.-C., Oh, B.-C., Shin, S., Kim, H.-J., Oh, T.-K., Kim, Y.-O. et al. (2000). Crystal structures of a novel, thermostable phytase in partially and fully calciumloaded states. Nature Struct. Biol. 7, 147–153. 33. Pons, T., Go´mez, R. & Chinea, G. (2003). b-Propellers: associated functions and their role in human diseases. Curr. Med. Chem. 10, 505–524. 34. Brady, G. P. & Stouten, P. F. W. (2000). Fast prediction and visualization of protein binding pockets with PASS. J. Comput. Aid. Mol. Des. 14, 383–401. 35. Morris, G. M., Goodsell, D. S., Halliday, R. S., Huey, R., Hart, W. E., Belew, R. K. & Olson, A. J. (1998). Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function. J. Comput. Chem. 19, 1639–1662. 36. Schiefner, A., Breed, J., Bo¨sser, L., Kneip, S., Gadel, J., Holtmann, G. et al. (2004). Cation–p interactions as determinants for binding of the compatible solutes glycine betaine and proline betaine by the periplasmic ligand-binding protein ProX from Escherichia coli. J. Biol. Chem. 279, 5588–5596. 37. Alberts, I. L., Nadassy, K. & Wodak, S. (1998). Analysis of zinc binding sites in protein crystal structures. Protein Sci. 7, 1700–1716. 38. Nurizzo, D., Turkenburg, J. P., Charnock, S. J., Roberts, S. M., Dodson, E. J., McKie, V. A. et al. (2002). Cellvibrio japonicus a-L-arabinanase 43A has a novel five-blade b-propeller fold. Nature Struct. Biol. 9, 665–668. 39. Dai, J., Liu, J., Deng, Y., Smith, T. M. & Lu, M. (2004). Structure and protein design of a human platelet function inhibitor. Cell, 116, 649–659. 40. Beisel, H.-G., Kawabata, S.-I., Iwanaga, S., Huber, R. & Bode, W. (1999). Tachylectin-2: crystal of a specific GlnNac/GalNac-binding lectin involved in the innate immunity host defense of the japanese horsehoe crab Tachypleus tridentatus. EMBO J. 18, 2313–2322. 41. Yamaguchi, A., Tada, T., Wada, K., Nakaniwa, T., Kitatani, T., Sogabe, Y. et al. (2005). Structural basis for thermostability of endo-1,5-a-L-arabinanase from Bacillus thermodenitrificans TS-3. J. Biochem. 137, 587–592. 42. Proctor, M. R., Taylor, E. J., Nurrizo, D., Turkenburg, J. P., Lloyd, R. M., Vardakou, M. et al. (2005). Tailored

43.

44.

45.

46.

47.

48.

49. 50.

51. 52. 53.

54.

55. 56.

57.

58.

catalysts for plant cell-wall degradation: redesigning the exo/endo preference of cellvibrio japonicus arabinanase 43A. Proc. Natl Acad. Sci. USA, 102, 2697–2702. Huang, K.-F., Liu, Y.-L., Cheng, W.-J., Ko, T.-P. & Wang, A. H.-J. (2005). Crystal structure of human glutaminyl cyclase, an enzyme responsible for protein N-terminal pyroglutamate formation. Proc. Natl Acad. Sci. USA, 102, 13117–13122. Ezura, Y., Kajita, M., Ishida, R., Yoshida, S., Yochida, H., Suzuki, T. et al. (2004). Association of multiple nucleotide variations in the pituitary glutaminyl cyclase gene (QPCT) with low radial BMD in adult women. J. Bone Miner. Res. 19, 1296–1301. Saido, T. C., Yamao-Harigaya, W., Iwatsubo, T. & Kuwashima, S. (1996). Amino- and carboxyl-terminal heterogeneity of b-amyloid peptides deposited in human brain. Neurosci. Letters, 215, 173–176. Harigaya, Y., Saido, T. C., Eckman, C. B., Prada, C. M., Shoji, M. & Younkin, S. G. (2000). Amyloid beta protein starting pyroglutamate at position 3 is a major component of the amyloid deposits in the Alzheimer’s disease brain. Biochem. Biophys. Res. Commun. 276, 422–427. Schilling, S., Hoffmann, T., Manhart, S., Hoffmann, M. & Demuth, H.-U. (2004). Glutaminyl cyclases unfold glutamyl cyclase activity under mild acid conditions. FEBS Letters, 563, 191–196. Azarkan, M., Clantin, B., Bompard, C., Belrhali, H., Baeyens-Volant, D., Looze, Y. et al. (2005). Crystallization and preliminary X-ray diffraction studies of the glutaminyl cyclase from Carica papaya latex. Acta Crystallog. sect. F, 61, 59–61. Kabsch, W. (1993). Processing of rotation diffraction data from crystals of initially unknown symmetry and cell constants. J. Appl. Crystallog. 26, 795–800. Hendrickson, W. A. & Teeter, M. M. (1981). Structure of the hydrophobic protein crambin determined directly from the anomalous scattering of sulfur. Nature, 290, 107–113. Ramagopal, U. A., Dauter, M. & Dauter, Z. (2003). Phasing on anomalous signal of sulfurs: what is the limit? Acta Crystallog. sect. D, 59, 1020–1027. Wang, B. C. (1985). Resolution of phase ambiguity in macromolecular crystallography. Methods Enzymol. 115, 90–112. Perrakis, A., Morris, R. & Lamzin, V. S. (1999). Automated protein model building combined with iterative structure refinement. Nature Struct. Biol. 6, 458–463. Bru¨nger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallog. sect. D, 54, 905–921. Kleywegt, G. J. & Jones, T. A. (1998). Databases in protein crystallography. Acta Crystallog. sect. D, 54, 1119–1131. Roussel, A. & Cambillau, C. (1991). The TURBOFRODO Graphics Package (Graphic, S., ed.) Silicon Graphics Geometry Partners Directory, vol. 81, Silicon Graphics, Mountain View, CA. Shimazaki, A., Makino, Y., Omichi, K., Odani, S. & Hase, S. (1999). A new sugar chain of the proteinase inhibitor from latex of Carica papaya. J. Biochem. 125, 560–565. Makino, Y., Shimazaki, A., Omichi, K., Odani, S. & Hase, S. (2000). Processing pathway deduced from the structures of N-glycans in Carica papaya. J. Biochem. 127, 1121–1126.

470

Papaya Glutaminyl Cyclase Crystal Structure

59. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallog. sect. D, 53, 240–255. 60. The, C. C. P. 4. (1994). CCP4 suite: programs for protein crystallography. Acta Crystallog. sect. D, 50, 760–763. 61. Vagin, A. A., Steiner, R. A., Lebedev, A. A., Potterton, L., McNicholas, S., Long, F. & Murshudov, G. N. (2004). Organization of prior chemical knowledge and guidelines for its use. Acta Crystallog. sect. D, 60, 2184–2195. 62. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 129–147. 63. Kabsch, W. & Sander, C. (1983). Dictionary of protein secondary structure: pattern recognition of hydrogenbonded and geometrical features. Biopolymers, 22, 2577–2637. 64. Hutchinson, E. G. & Thornton, J. M. (1996). PROMOTIF, a program to identify and analyze structural motifs in proteins. Protein Sci. 5, 212–220.

65. Honig, B. & Nicholls, A. (1995). Classical electrostatics in biology and chemistry. Science, 268, 1144–1149. 66. Weiner, S. J., Kollman, P. A., Case, D. A., Singh, U. C., Ghio, C., Alagona, G. et al. (1984). A new force field for molecular mechanical simulation of nucleic acids and proteins. J. Am. Chem. Soc. 106, 765–784. 67. Hete´nyi, C. & van der Spoel, D. (2002). Efficient docking of peptides to proteins without prior knowledge of the binding site. Protein Sci. 11, 1729–1737. 68. Kraulis, P. J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallog. 24, 946–950. 69. Esnouf, R. M. (1999). Further additions to MolScript version 1.4, including reading and contouring of electron-density maps. Acta Crystallog. sect. D, 55, 938–940. 70. Merritt, E. A. & Murphy, M. E. P. (1994). D version 2.0—a program for photorealistic molecular graphics. Acta Crystallog. sect. D, 50, 869–873. 71. Wallace, A. C., Laskowski, R. A. & Thorton, J. M. (1995). Ligplot: a program to generate schematic diagrams of protein–ligand interactions. Protein Eng. 8, 127–134.

Edited by I. Wilson (Received 12 July 2005; received in revised form 8 December 2005; accepted 9 December 2005) Available online 4 January 2006