Queen Bee Pheromone Binding Protein pH-Induced Domain Swapping Favors Pheromone Release

doi:10.1016/j.jmb.2009.05.067

J. Mol. Biol. (2009) 390, 981–990

Available online at www.sciencedirect.com

Queen Bee Pheromone Binding Protein pH-Induced Domain Swapping Favors Pheromone Release Marion E. Pesenti 1,2 , Silvia Spinelli 1,2 , Valérie Bezirard 3 , Loïc Briand 3 , Jean-Claude Pernollet 3 , Valérie Campanacci 1,2 , Mariella Tegoni 1,2 ⁎ and Christian Cambillau 1,2 ⁎ 1

Architecture et Fonction des Macromolécules Biologiques, UMR 6098 CNRS, 13288 Marseille, France 2

Universités d'Aix-Marseille, 13288 Marseille, France 3

INRA, UMR 1197, Neurobiologie de l'Olfaction et de la Prise Alimentaire, Biochimie de l'Olfaction et de la Gustation, F-78352 Jouy-en-Josas, France Received 14 April 2009; received in revised form 19 May 2009; accepted 22 May 2009 Available online 28 May 2009

In honeybee (Apis mellifera) societies, the queen controls the development and the caste status of the members of the hive. Queen bees secrete pheromonal blends comprising 10 or more major and minor components, mainly hydrophobic. The major component, 9-keto-2(E)-decenoic acid (9ODA), acts on the workers and male bees (drones), eliciting social or sexual responses. 9-ODA is captured in the antennal lymph and transported to the pheromone receptor(s) in the sensory neuron membranes by pheromone binding proteins (PBPs). A key issue is to understand how the pheromone, once tightly bound to its PBP, is released to activate the receptor. We report here on the structure at physiological pH of the main antennal PBP, ASP1, identified in workers and male honeybees, in its apo or complexed form, particularly with the main component of the queen mandibular pheromonal mixture (9-ODA). Contrary to the ASP1 structure at low pH, the ASP1 structure at pH 7.0 is a domain-swapped dimer with one or two ligands per monomer. This dimerization is disrupted by a unique residue mutation since Asp35 Asn and Asp35 Ala mutants remain monomeric at pH 7.0, as does native ASP1 at pH 4.0. Asp35 is conserved in only ∼ 30% of mediumchain PBPs and is replaced by other residues, such as Asn, Ala and Ser, among others, thus excluding that they may perform domain swapping. Therefore, these different medium-chain PBPs, as well as PBPs from moths, very likely exhibit different mechanisms of ligand release or receptor recognition. © 2009 Elsevier Ltd. All rights reserved.

Edited by I. Wilson

Keywords: honeybee; Apis mellifera; pheromone binding protein; crystal structure; signal transduction

Introduction Olfaction is a very elaborate sense since the olfactory system has to detect thousands of volatile molecules at low concentrations and discriminate between some of them differing by only one or a few *Corresponding authors. Architecture et Fonction des Macromolécules Biologiques, UMR 6098 CNRS, Case 932, 163 Avenue de Luminy, 13288 Marseille Cedex 09, France. E-mail addresses: [email protected]; [email protected]. Abbreviations used: PBP, pheromone binding protein; OBP, odorant binding protein; 9-ODA, 9-keto-2(E)decenoic acid; nBBSA, n-butyl-benzenesulfonamide; OR, odorant receptor; SNMP, sensory neuron membrane protein.

atoms. In insects, the first event in pheromone detection is the capture of the small volatile molecule by a pheromone binding protein (PBP).1 It was for long a matter of discussion whether these proteins would be simple solubilizers/carriers or would play a more sophisticated role. A series of studies has recently documented the essentiality of PBPs in signal transduction and assessed their essential physiological role.2–5 It has been proposed that after the first step of PBP/pheromone association, the PBP interacts with a sensory neuron membrane protein (SNMP),6,7 a co-receptor8 associated with the olfactory receptor heterodimer.9 Each monomer of the odorant receptor (OR) dimer possesses seven transmembrane helices, but contrary to mammals, they are not G-protein-coupled receptors. Two recent studies have revealed that ORs in fact belong to a new class of ligand-gated ion channels10 but

0022-2836/$ - see front matter © 2009 Elsevier Ltd. All rights reserved.

982 may also involve G-protein activation.11 Furthermore, it has been shown in the case of Drosophila Lush that although PBPs are able to bind different nonspecific ligands with equal affinity as for the specific pheromone,12–14 only the specific pheromone (vaccenyl acetate) can trigger the appropriate conformational change, allowing it to be recognized by the specific receptor.5 To date, only the first step of this cascade has been investigated structurally. Several PBP structures, alone or in complex with ligands, have been described. The PBP of Bombyx mori (BmorPBP) in complex with bombykol was solved first.15 Together with the structure of the apo-protein solved by NMR,16,17 this structure led the authors to propose a mechanism in which pheromones would be bound at neutral pH, while a seventh helix would fold in the cavity and expel the pheromone at acidic pH. They proposed that the lower pH at the membrane level (by 1.5–2.5 U) would be the trigger of ligand release at the vicinity of the membrane, close to the OR. The finding of an apo-PBP structure at neutral pH without an internal helix somewhat complicated the interpretation.18 It was then proposed that an interplay of ligand affinity and pH would orchestrate a cycle of ligand loading in the sensillum lymph and release near the ORs in the ciliary membranes of olfactory sensory neurons. Besides the class of long-chain BmorPBP-like PBPs (∼ 140 amino acids long), two other classes of six cysteine PBPs have been described: the mediumchain-length PBPs (∼ 120 amino acids long) and the short-chain-length PBPs (∼100 amino acids long). Three structures of PBPs belonging to the mediumchain-length class have been solved alone and/or in the presence of a ligand: Drosophila melanogaster Lush, 19,20 Apis mellifera ASP114 and Anopheles gambiae AgamPBP. 21 Finally, the PBP structure from cockroach, Leucophaea maderae (LmaPBP), is the unique example of the third class.13 Whereas the C-terminus of long-chain PBPs is inserted in apoPBP structures as a seventh helix, the C-terminus of medium-chain PBPs is inserted as an extended stretch, forming a wall of the cavity, even in the presence of ligand.19,22 In the three structures, these C-termini are locked by a double hydrogen/ionic bond between the carboxylic end of the last residue and two other amino acids—one bearing a positive charge (His, Lys and Arg, respectively) and a tyrosine. In the case of AgamPBP, Wogulis et al.21 proposed that this structure might be stable only at neutral pH and that the C-terminus might be expelled at low pH due to the pH-induced rupture of the ionic bond, which would then destroy the binding site and allow ligand release. They therefore proposed that long- and medium-chain PBPs might share a similar mechanism of ligand expulsion triggered by the low pH near the sensory neuron membrane. We have previously investigated structural aspects of ligand binding of ASP1, a medium-length PBP from honeybee, using three ligands—the main component of queen bee pheromone [9-keto-2(E)-dece-

Honeybee PBP ph-Induced Domain Swapping

noic acid (9-ODA)], a serendipitous ligand [n-butylbenzenesulfonamide (nBBSA)] and a fatty acid.22 These studies were performed at three pHs, 4.0, 5.5 and 7.0; they have demonstrated that (i) the Cterminus lock is not destroyed even at low pH, (ii) the C-terminus always remains within the protein core and (iii) ligands are bound to PBP at pH 4.0 and pH 5.5 but are expelled at neutral pH due to an electrostatic clash involving Asp35. Here, studies in solution and crystallization at pH 7.0 reveal the formation of a domain-swapped dimer of ASP1 at this pH that is able to bind ligands, but with lower affinity compared with pH 4.0. Mutations of Asp35 in Asn favoring hydrogen bonding at all pHs or in Ala (neutral) led to the identification of a third crystal form at pH 7.0 and demonstrated an increased affinity of one of these mutants for a ligand at this pH. Based on the observation that Asp35 is not conserved, we propose that different mechanisms of ligand binding and release may coexist for the different members of the medium-chain-length PBPs, which are therefore not triggered by the membrane vicinity but most probably by interactions with cell surface protein receptors or co-receptors.

Results Native ASP1 ligand binding followed by fluorescence quenching We have previously shown that crystals of a complex between ASP1 and nBBSA obtained at pH 5.5 retained the ligand when soaked at pH 4.0, while soaking at pH 7.0 expelled the ligand from the crystal.22 In order to confirm these results in solution, we decided to perform binding experiments following tryptophan quenching, a technique useful to assess binding of ligands to PBPs, since the Cterminal segment of ASP1 bears a Trp residue at position 116.12 In our structures, this residue is in contact with or is very close (less than 5 Å) to the ligands; therefore, it is a good candidate as a reporter of ligand binding. Using this method, we have measured the affinity of ASP1 for nBBSA and 9ODA. The compounds studied, 9-ODA and nBBSA, can be solubilized in organic solvents, such as methanol, ethanol and DMSO, which are classic solvents used for such studies.12 However, attempts to use these solvents produced an important quenching of the tryptophan, masking the ligand effect. We thought that this effect might be due to the binding of these strong hydrogen donor alcohols. Alternatively, it cannot be excluded that this effect might be due to the presence of phthalate contaminants in alcohol, as reported elsewhere.23 We therefore used a weak hydrogen donor solvent, t-butyl alcohol, which did not interfere with the binding site, while efficiently solubilizing our compounds. At pH 4.0, the addition of either compound yields a decrease of the Trp fluorescence by up to 50%

983

Honeybee PBP ph-Induced Domain Swapping Table 1. Dissociation constants of nBBSA and 9-ODA with native ASP1 and D35N mutant Ligand/Mutant nBBSA 9-ODA nBBSA 9-ODA nBBSA/D35N

pH

Kd1 (nM)

Kd2 (μM)

4 4 7 7 7

7.0 ± 2.0 8.5 ± 2.5 55.0 ± 8.0 60.0 ± 15.0 23 ± 4

6.5 ± 1.5 5.0 ± 2.0 33.2 ± 3.0 nd 9.2 ± 7

intensity and a displacement of its maximum from 342 down to 335 nm (Supplementary Fig. 1). A first quasi-saturation was obtained at ∼ 0.25 μM (equivalent to the protein concentration), but quenching continued up to 10 μM, where the second saturation was observed. In this latter phase, the maximum fluorescence remained constant at 335 nm. The first constants, as evaluated by GraphPad, are 8.5 ± 3 and 7.0 ± 2 nM for 9-ODA and nBBSA, respectively (Table 1). The second constants are ∼ 1000-fold larger, 5.0 ± 2 and 6.5 ± 1.5 μM for 9-ODA and nBBSA, respectively. The second constants probably report a nonspecific association with another protein pocket, and this second binding site has never been observed in any crystal structure. Given the results of crystal soaking at pH 7.0 on ligand binding, leading to ligand dissociation, we were surprised to observe tryptophan fluorescence quenching at this pH (Supplementary Fig. 1). At pH 7.0, the addition of either compound yields a decrease of the Trp fluorescence by up to 50% intensity for nBBSA but only that of 20% for 9-ODA. The displacement of the peak maxima is also observed, from 339 ± 1 down to 337 ± 1 nm. As observed at pH 4.0, a first saturation was obtained with nBBSA at ∼ 0.25 μM (equivalent to the protein concentration), but quenching continued up to 10 μM, where the second saturation was observed. In this latter phase, the maximum of fluorescence remained constant at 337 ± 1 nM. The first constants, as evaluated by GraphPad, are 60.0 ± 15 and 55.0 ± 8 nM for 9-ODA and nBBSA, respectively (Table 1). The second constant for nBBSA is ∼ 500fold larger, 33.0 ± 3 μM. The second constants probably reflect the association of the second molecule in the protein cavity (see below). When considering that the protein concentration is 250 nM, the first Kd values, 30 times lower than the protein concentration, are of course overestimated. The affinity of ASP1 for 9-ODA and nBBSA compounds should therefore be somewhat lower than that indicated by the calculated Kd values of 8–9 nM. Unfortunately, lower protein concentrations were not usable for Kd determination. Structure of ASP1 crystallized at pH 7.0: the domain-swapped structure The discrepancy between our observations in crystals and those in solution prompted us to crystallize the native protein at pH 7.0. No crystal could be obtained of form 1 above pH 5.9, however. In a wide screening for new crystallization condi-

tions, we obtained a new crystal form at pH 7.0 (form 2). Crystals were obtained in the apo form and in complex with nBBSA and 9-ODA (Table 2). The ASP1 structure in crystal form 2 (Fig. 1a and b) differs widely from the structures, apo or liganded, of ASP1 in crystal form 1 (Fig. 1b). The core of the protein is similar between residues 13 and 110, with a root-mean-square deviation (rmsd) of 1.28 Å, but the N- and C-termini follow different tracks, with rmsd values of 4.32 and 1.51 Å, respectively, for the two forms complexed with 9-ODA (Fig. 1a). ASP1 forms dimers in which the Nterminus (residues 8–13) of one monomer forms a parallel β-sheet with the C-terminus (residues 114– 118) of the other monomer, resulting in a domainswapped structure stabilized by five hydrogen bonds (Fig. 1b). There are 36 residues implicated in the interaction, involving a buried surface of 1350 Å2, with the C-terminus still inserted in the cavity. Superposition of two dimers with their subunits A and B remarkably shows the lack of 2-fold symmetry of the dimer, with a rotation angle between the nonaligned subunits of ∼ 12° (Fig. 2a). As a consequence, the two subunits differ in many points: (i) Subunit A contains one ligand (nBBSA or 9-ODA) and has a volume smaller than that of subunit B, which contains two ligands (nBBSA and 9-ODA; Supplementary Fig. 2). The ligand in subunit A has a position close to that of the ligand in crystal form 1. For nBBSA, a mean shift of 4 Å is observed between the two forms (Fig. 1b). In contrast with the holoprotein, apo-ASP1 possesses two empty cavities of similar volume. (ii) The C-terminus lock still exists but is different between the two subunits (Supplementary Table 1). It involves the COO− group of Ile119, hydrogen bound to Lys17A (as in form 1) in monomer A (apo form only) and in monomer B (all forms; Supplementary Table 2), an interaction with Asp11 COO− via a water molecule in the complexed monomer A and, lastly, weak hydrogen bonds with Val13 NH in the complexes of monomer A. In monomer B, only the hydrogen bound to Lys17A is observed, with Val13B being farther away due to the larger cavity containing two ligands (Supplementary Table 1). None of the ligand molecules belonging to subunit A or B interacts strongly with ASP1. All interactions involve low-energy van der Waals contacts and no hydrogen bond, in contrast with what is observed in form 1. The second ligand in the cavity of subunit B is closer to the cavity opening than the first one. The most buried ligand has an overall position comparable with the ligand in subunit A and that in form 1 but with looser contacts with the binding pocket. Structure of the ASP1 mutants D35N and D35A and the effect of pH In order to confirm the role of Asp35 in ligand ejection at pH 7.0 and its effect on domain swapping, we have produced a D35N mutant that can establish hydrogen bonds with Val118 indepen-

984

Table 2. Data collection and refinement statistics D35N nBBSA

Data collection Crystal form PBP access code Space group Unit cell parameters (Å) a b c Beamline Detectors Wavelength (Å) Rotation range (°) Resolution range (Å) No. of observations No. of unique reflections Completeness Redundancy I/σI Rsym (%) Refinement Oligomerization Resolution range (Å)

D35A nBBSA

9-ODA

nBBSA

pH 5.5

pH 7.0 soak

pH 4.0 soak

pH 7.0

pH 7.0

pH 5.5 soak

2 3CZ2 P212121

2 3CYZ P212121

2 3CZ1 P212121

1 3D75 C2221

1 3D76 C2221

1 3D77 C2221

3 3D78 P21212

3 3D73 P21212

3 3D74 P21212

35.8 75.3 84.2 ID23-EH1 ADSC Q315R 0.886 180 50.00–2.50 (2.64–2.50) 52,583 (7875) 8426 (1214) 99.9 (100) 6.2 (6.5) 14.6 (5.1) 11.5 (31.3)

36.3 75.5 84.1 ID29 ADSC Q315R 0.979 180 84.21–1.80 (1.90–1.80) 147,798 (18,021) 21,305 (2614) 96.4 (84.1) 6.9 (6.9) 16.9 (3.7) 9.1 (45.7)

35.4 75.6 84.1 ID23-EH1 ADSC Q315R 0.954 200 35.00–1.50 (1.58–1.50) 141,978 (20,849) 36,737 (5313) 99.4 (100) 3.9 (3.9) 12.7 (3.7) 7.8 (30.1)

78.5 83.7 47.9 Home MAR345dtb 1.542 90 30.36–2.30 (2.42–2.30) 25,819 (3723) 7140 (1040) 98.4 (99.8) 3.6 (3.6) 16.5 (3.1) 5.6 (39.0)

79.0 84.2 47.6 Home MAR345dtb 1.542 160 25.13–1.90 (2.00–1.90) 77,804 (10,989) 12,833 (1837) 99.8 (100) 6.1 (6.0) 24.8 (6.5) 5.6 (32.2)

79.2 84.0 47.5 ID29 ADSC Q315R 0.979 180 57.64–1.70 (1.79–1.70) 119,393 (17,751) 17,650 (2545) 99.1 (99.6) 6.8 (7.0) 26.4 (4.5) 4.7 (34.9)

62.1 60.9 57.3 ID23-EH1 ADSC Q315R 0.984 125 50.00–1.55 (1.63–1.55) 141,782 (19,220) 32,103 (4588) 99.6 (99.6) 4.4 (4.2) 20.3 (3.8) 4.8 (32.1)

61.9 60.9 56.7 Home MAR345dtb 1.542 160 30.43–2.01 (2.12–2.01) 88,730 (10,192) 14,418 (1734) 97.4 (82.2) 6.2 (5.9) 22.2 (6.4) 5.7 (25.4)

61.1 60.8 56.3 Home MAR345dtb 1.542 100 30.39–2.10 (2.21–2.10) 49,058 (6924) 12,708 (1804) 99.8 (100) 3.9 (3.8) 14.5 (4.3) 6.6 (29.0)

Swapped dimer 35.0–2.50 (2.56–2.50) 7568 (561)

Swapped dimer 20.00–1.80 (1.85–1.80) 20,184 (1107)

Swapped dimer 20.00–1.50 (1.54–1.50) 35,532 (2613)

Monomer 30.00–2.30 (2.36–2.30) 6435 (466)

Monomer 15.00–1.90 (1.95–1.90) 11,728 (778)

Monomer 30.00–1.70 (1.74–1.70) 16,595 (1204)

Crystal dimer 15.00–1.60 (1.64–1.60) 28,039 (2026)

Crystal dimer 20.00–2.03 (2.08–2.03) 13,343 (892)

Crystal dimer 15.00–2.10 (2.15–2.10) 11,717 (823)

1856 320 39

1876 303 42

907 56 14

910 155 14

947 140 14

1905 266 42

1846 154 42

1845 122 42

0.162/0.199 0.196/0.204

0.152/0.198 0.165/0.220

0.179/0.221 0.224/0.231

0.190/0.241 0.499/0.583

0.161/0.191 0.162/0.212

0.159/0.206 0.151/0.221

0.175/0.224 0.151/0.221

0.175/0.229 0.229/0.254

0.013 1.413

0.013 1.601

0.013 1.940

0.013 1.848

0.014 2.211

0.013 1.498

0.013 1.375

0.012 1.469

20.5 32.5 52.2

14.8 28.7 30.6

42.4 47.3 72.8

33.0 44.7 60.0

30.5 44.6 44.6

27.2 36.6 22.1

34.8 37.8 34.7

35.8 38.1 34.5

92.7 7.3

92.2 7.8

90.4 9.6

93.2 6.8

94.2 5.8

93.3 6.7

92.9 7.1

93.8 6.2

Unique reflections No. of atoms Protein 1811 Water/Buffer 30 Ligands — Rcryst/Rfree All 0.220/0.278 Last shell 0.284/0.366 rmsd Bonds (Å) 0.012 Angles (°) 1.213 Mean B-value (Å2) Protein 15.3 Water/Ions 23.4 Ligands — PROCHECK Ramachandran plot (%) Most favored region 89.9 Additionally allowed regions 10.1

Values in brackets are those concerning the last shell.

Honeybee PBP ph-Induced Domain Swapping

Apo

Honeybee PBP ph-Induced Domain Swapping

985

Fig. 1. Ribbon view of ASP1 crystal forms. (a) Form 2 in complex with 9-ODA obtained at pH 7.0. Each monomer is colored from blue (N-terminus) to red (C-terminus). The N- and C-termini form a domain-swapped structure. (b) Form 2 in complex with nBBSA obtained at pH 7.0 (yellow, A; purple, B; spheres, nBBSA) is superimposed to form 1 in complex with nBBSA at pH 5.5 (red; red grid, nBBSA). The displacement of the N-terminus from form 1 to form 2 is indicated by the thin blue arrow. Inset: Stick representation of the parallel β-sheet established between the N-terminus of monomer A and the C-terminus of monomer B. Note the conformational changes (besides domain swapping) of helix 2.

dently of pH. We have crystallized this mutant in the conditions used to yield form 1 at pH 5.5. Crystals obtained at pH 5.5 were then soaked either at pH 4.0 or at pH 7.0. The three structures were then determined and were found to be indistinguishable, with

rmsd values within experimental errors (0.2–0.35 Å). The structures resemble closely that of the native ASP1 in complex with nBBSA at acidic pH, and the rmsd value between their Cα atoms is 0.5 Å. The ligand however remains bound at pH 7.0, in contrast

Fig. 2. Conformational changes of native ASP1 and mutants. (a) Ribbon view of the superposition of monomers A and B of two ASP1 dimers. The dimer is asymmetrical (tilt angle of 12°). (b and c) The C-termini lock in monomer A (b) and monomer B (c). Note the different H-bonding schemes and the different positions of the ligands. (d–f) Superpositions involving Asp35, the C-terminus and the nBBSA ligand. (d) D35N mutant (form 1, pH 7.0 soaked) with native ASP1 soaked at pH 4.0 (form 1). (e) D35N mutant at pH 5.5 (form 1) with the structures obtained at pH 4.0 and pH 7.0 (form 1). (f) D35N mutant at pH 7.0 (form 3) with D35A at pH 7.0 (form 3); note the displacement of Ala35.

986 to what is observed with the native protein. The C = O moiety of Asn35 is directed toward the N–H group of Val118 with a distance of 2.63 Å, while the NH2 group of Asn35 forms a hydrogen bond with the C = O group of Val118 with a distance of 2.93 Å, in a similar fashion as form 1 at acidic pH (Fig. 2d and e). Following the same strategy as for the native protein, we have tried to crystallize the D35N mutant at pH 7.0. New conditions of crystallization have been identified with the apo-protein mutant, leading to a new crystal form (form 3) different from that of the native protein at pH 5.5 or pH 7.0. Despite having a crystallographic dimer in the asymmetric unit, ASP1 form 3 is clearly monomeric because the dimer interaction surface is only 500 Å2 and no domain swapping is observed. This value is very close to that observed at pH 5.5 for crystal contacts with symmetrically related molecules for ASP1 form 1 (440 Å2) and for ASP1 form 3 (430 Å2). The two subunits are, overall, close together (rmsd = 0.54 Å) and to D35N form 1 crystals (rmsd = 0.8 ± 0.1 Å), with some deviations. While subunit A has one ligand, subunit B has a ligand in a similar position in the cavity plus a second that is weakly defined and bound close to the entrance of the cavity. This second ligand provokes a slight opening of this mouth, by ∼ 3 Å, at the level of loop 70–75. Comparing new D35N form 3 crystals with those of form 1, we noticed a displacement of loop 108–112, preceding the C-terminal segment. This loop was however often poorly defined in form 1 crystal structures. Despite the different conformation of segment 108–112, the C-terminus (113–119) has the same conformation as that in the native ASP1 at acidic pH and as the mutants in form 1 (Supplementary Table 1). The C-terminus is stabilized by two hydrogen bonds between the Ile119 COO− group and Tyr48 OH and Lys17 NH3+, as in native ASP1 form 1 at acidic pHs. It is therefore different from the structure of the apo form soaked at pH 7.0. Asn35 establishes two tight hydrogen bonds with Val118, similarly as in the soaked D35N crystals belonging to form 1. Then, Asp35 was replaced with a neutral residue, Ala, a side chain providing neither attraction nor repulsion at any pH used to crystallize the protein or to soak the crystals. The D35A mutant in complex with nBBSA was crystallized at pH 7.0, leading to the same crystal form (form 3) as for the D35N. Both structures are very close, with an rmsd of 0.28 Å. The structure of this complex is very similar to those obtained with the D35N mutant, and the same structural features are observed. The Cβ of the Ala residue is slightly shifted toward the protein surface by ∼ 1.5 Å (Fig. 2f). Subsequently, one of these D35A form 3 crystals was soaked at pH 5.5 and data were collected at 2.10-Å resolution. The pH 5.5 structure remains very close to that obtained at pH 7.0 (rmsd = 0.16 Å). In conclusion, neither the pH nor the liganded state has any effect on the conformation of the C-terminus in both mutants, D35N and D35A, contrary to the native protein.22

Honeybee PBP ph-Induced Domain Swapping

ASP1 D35N mutant ligand binding followed by fluorescence quenching We tested binding of nBBSA to ASP1 mutant D35N using Trp fluorescence quenching at pH 7.0. Addition of nBBSA produces a large decrease of Trp fluorescence accompanied by a shift of its maximum, in contrast with the results obtained with the wild-type protein and similar to the wild-type protein behavior at pH 4.0 (Supplementary Fig. 3). Analysis of the data using GraphPad yielded a Kd value of 23 ± 4 nM; this value is three times higher than that observed for the native protein at pH 4.0 but is also three times lower than that for the native protein at pH 7.0, thus confirming the stabilizing effect of the mutation on the C-terminus (Table 2).

Discussion We have measured affinity constants for ASP1 toward its pheromone, 9-ODA, in a Kd range of 7– 60 nM, depending on pH. Although these values are slightly underestimated due to technical constraints (see Results), it should be mentioned that the binding constants found here at pH 4.0 are significantly lower than those found for pheromones in moths at pH 7.0—for example, in Antheraea polyphemus) by Kaissling (Kd = 60 nM),24 by Du and Prestwich (Kd = 640 nM)25 and by Campanacci et al. (Kd = 200 nM);12 in B. mori by Leal et al. (Kd = 105 nM;26 and in Mamestra brassicae by Campanacci et al. (Kd = 510 nM).12 This may be due in some cases to the use of different techniques or experimental procedures: the constants for A. polyphemus, for example, span over 1 order of magnitude range depending on the authors. Furthermore, the values measured for ASP1 at pH 7.0 are closer to those observed in moths. It however seems that ASP1 has intrinsically smaller Kd values toward 9-ODA and nBBSA compared with moth, especially at low pH. We have shown in a previous study that the ASP1 active-site shape is not maintained at neutral pH and that protonation of Asp35 side chain is necessary to maintain the C-terminus in contact with the core of the protein. Comparison of the structure of ASP1 apo form at pH 5.9 with that of the complexes at pH 5.5 has indicated that the presence of a ligand helps maintaining the C-terminus in a conformation stacked against the protein core. Soaking of crystals of the nBBSA/ASP1 complex at pH 7.0 resulted in a conformational change of the C-terminus expelling the ligand and yielding the ASP1 apo form. Our fluorescence quenching experiments confirm that tight binding occurs at pH 4.0. However, in contrast with our expectations, binding is also observed at pH 7.0, but with a 10-fold affinity decrease. Because we thought that the crystal lattice may block a larger conformational change of ASP1 when soaked at pH 7.0, we performed crystallization at pH 7.0. This yielded a domain-swapped dimeric form never observed before with PBPs. PBP dimerization was for long a matter of controversy: it was sometimes

Honeybee PBP ph-Induced Domain Swapping

observed in solution and postulated in the crystal, but with the lack of any conserved dimer arrangement, these contacts were suggested to be opportunistic. Notably, bovine odorant binding protein (OBP), with a totally different fold as compared with insect PBP/OBP, has been shown to also exhibit a domain-swapped structure with two differently shaped subunits.27 The ASP1 pH 7.0 soaked crystal structure demonstrates clearly that soaking experiments have loosened the C-terminus and detached it from the protein core. We postulate that in solution at pH 7.0 the loose C-terminus can recruit the N-terminus of another ASP1 molecule to form a more stable domain-swapped dimeric form (Fig. 3). This form is still able to bind ligands, but with lower affinity than the acidic ASP1 monomeric form, as demonstrated in solution and in the crystal. The trigger of this mechanism should be Asp35, as shown by the mutation experiments of D35N and D35A mutants, which exhibit exactly the same monomeric structure whatever their pH and binding state are (Figs. 2 and 3). It should be noticed that the mutant structures make it possible to reject the hypothesis of a destabilization of the N-terminus at pH 7.0 in the formation of the swapped dimer since it is not formed with neither of them. Surprisingly, our observations are at odds with the mechanism proposed for the “long” PBPs, such as those from B. mori.15,16,28–30 For those PBPs, the C-terminus adopts a helical structure and folds into the PBP internal cavity at acidic pH, preventing ligand access or, eventually, pushing out a bound ligand outside the cavity. At pH 7.0, in contrast, the

987 C-terminus stretches outside the cavity, lining on the PBP surface, and the binding site is accessible. Furthermore, fluorescence quenching experiments have demonstrated that while BmorPBP is able to bind bombykol only at neutral pH and not at pH 4.0, a mutant in which the terminal segment is deleted binds bombykol at both pHs.26 These observations have led the authors to propose that the mechanism of ligand release may be triggered by low pH, a situation that may be found close to the cell membranes, where the pheromone receptor is embedded. It was later proposed that this model might be general for all PBPs, based on observations on the structure of AgamPBP and on that of Lush,19 which both belong to the class of short C-termini PBPs as ASP1. In both cases, the moiety of the last carboxylic acid residue establishes H-bonds with His23 and Tyr54 (AgamPBP) and with Arg14 and Tyr47 (Lush). Because AgamPBP does not bind a nonspecific ligand, bombykol, at acidic pH, it was postulated that low pHs may disrupt the hydrogen bonds of the last C-terminal carboxylic acid residue with its partners and lead to its exit from the cavity, destroying the shape of the binding site. Our results with the D35N mutant clearly rule out this hypothesis. Residue 35 in ASP1 corresponds to residue 41 in AgamPBP, which is a Ser. Both Asn and Ser are polar noncharged residues able to establish hydrogen bonds with the C-terminal main chain independently of pH. The identification and importance of Asp35 as a crucial residue for shaping the binding site can be extended to other medium-chain-length PBPs. Blast

Fig. 3. Schematic representation of the effect of pH on native ASP1 and mutants. Top: Native ASP1 at acidic pH (pH 5.5; left) gives rise to a transient monomeric species at pH 7.0 (hypothesis) that leads to a predominant domainswapped species (right). Bottom: D35N (or D35A) mutant at pH 5.5 (left) remains unchanged at pH 7.0 (right).

988 results have identified 49 sequences of mediumchain-length PBPs with e-values below 3e− 06. In these PBPs, amino acid positions corresponding to those of Asp35 in ASP1 bear a large variety of residues that are charged, polar and hydrophobic (Supplementary Fig. 4). These side chains can be classified into 3 classes: those that form hydrogen bonds at acidic pH and are sensitive to neutral pH (Asp, Glu; 15 PBPs), those that form hydrogen bonds independently of pH (Ser, Asn; 25 PBPs) and those that do not form hydrogen bonds (Val, Ile, Ala; 4 PBPs). Five PBPs bear Arg or Lys. In a recent article, Laughlin et al.5 reported that in the apo- or nonspecifically bound PBP form, an ionic bond between Lush Asp118 and Lys87 blocks loop 116–120 in an inactive conformation. Upon binding with vaccenyl acetate, its specific pheromone, this bond is disrupted and the loop adopts the active conformation capable of triggering the activation of T1 neurons. A constitutively active PBP in which Asp118 was mutated in Ala was produced. Is this situation comparable with that of ASP1? Asp118 and Lys87 are not conserved in ASP1, where Pro113 and Ser84 are found at the same positions (Supplementary Fig. 5). However, Asp114 is in some structures at an ionic bond distance of Arg81 (Supplementary Fig. 4). Do we have a topological replacement of the ionic lock found in Lush, and does it follow the same structural pattern? A compilation of the distances between Asp114 and Arg81 is somewhat confusing: In crystal form 1, the ionic bond is formed only in the pheromone 9-ODA complex. With the D35N or D35A mutants, ionic bonds are formed in crystal form 3 but not in crystal form 1. Besides the fact that 9-ODA/ASP1 seems to act oppositely to vaccenyl acetate/Lush, no definitive conclusion can be drawn from these observations. This leads us to postulate that these different PBPs, as well as PBPs from moths, very likely exhibit different mechanisms of ligand release. It may occur in some cases at low pH and in other cases at neutral or high pH. It is therefore very unlikely that a unique factor, the pH drop at the vicinity of the membrane, might be the trigger of PBP ligand release. The recent and important discovery that PBPs may interact with their SNMP co-receptor6,7 offers an appealing alternative explanation. SNMP is a CD36related protein containing two transmembrane segments near the N- and C-termini with an ectodomain of ∼ 420 amino acids. Such an ectodomain should emerge out of the membrane surface and impose its own electrostatic influence on PBP to capture it and lead to pheromone release at the vicinity of the OR. The trigger of the ligand release might be the ionization state of the SNMP ectodomain, which may differ in different organisms. This ionization state may create a local pH, in some cases acidic and in other cases neutral or basic, triggering the conformational changes of the PBPs and the passage of its ligand to the SNMP. Alternatively, we cannot rule out a specific conformational change, yet undeciphered, as observed for vaccenyl acetate/ Lush.5

Honeybee PBP ph-Induced Domain Swapping

Materials and Methods Compounds 9-ODA was purchased from SPECS–Research Compound (Cumberland, MD); t-butyl-alcohol, from Fluka Chemie–Sigma-Aldrich (Munich, Germany). ASP1 expression and purification ASP1 was expressed recombinantly in Pichia pastoris and purified by reversed-phase chromatography as previously described.31 The purified protein was concentrated (Vivaspin 5000) in 10 mM Hepes, pH 6.0, to concentrations between 19.9 and 41.9 mg/ml, which were determined by spectrophotometry using the theoretical E280 = 15,580 M− 1 cm− 1. Native ASP1 appears as a dimer at pH 7.0 on a calibrated gel-filtration column, as described previously.31 D35N and D35A mutants were purified by a similar procedure but, in contrast with the native protein, yielded monomeric species on a calibrated gel-filtration column, as does the native protein at acidic pH (pH 5.0). Crystallization of ASP1 at pH 7.0 Initial crystallization screening was performed with the commercial kit SM1. The nano-L crystallization experiments were performed with a Cartesian HoneyBee X8 robot using the sitting-drop method in Greiner plates.32,33 Larger crystals of the apo form and different complexes (9-ODA and nBBSA) were obtained with the hanging-drop method using Limbro plates. The crystallization droplet contained 4 μl of a 26.8- to 41.9-mg/ml protein solution in 10 mM Hepes, pH 6.0, mixed with 1 μl of well solution containing 0.2 M magnesium chloride, 0.1 M Tris and 15%–17% polyethylene glycol 8000 at pH 7.0. Crystallization of mutants D35A and D35N of ASP1 The crystallization drops of the D35A and D35N mutants at pH 7.0 contained 200 nl of a 24-mg/ml protein solution in 10 mM Hepes, pH 7.0, mixed with 100 nl of well solution containing 2.1 M ammonium sulfate, 0.2 M diammonium phosphate and 20 mM disodium phosphate at pH 7.0. Crystals of 100 μm × 100 μm × 50 μm appearedafter 1 week. For soaking experiments, crystals were soaked in 500 nl of a solution identical with well solution but at pH 5.5. Crystals were cryo-cooled at 100 K using as cryoprotectant synthetic buffer liquor with 11.1%–14.3% glycerol. The crystallization drop of the D35N mutant at pH 5.5 contained 300 nl of a 22.0 mg/ml protein solution in 10 mM sodium citrate and 1.14 mM nBBSA, pH 5.5, mixed with 100 nl of well solution containing of 1.3 M ammonium sulfate, 67 mM sodium citrate, 33 mM SPG (succinic acid, sodium dihydrogen phosphate, glycine) buffer and 8.3% polyethylene glycol 1500 at pH 5.5. Crystals of 100 μm × 70 μm × 50 μm appeared after 1 week. For soaking experiments, crystals were soaked in 500 nl of a solution identical with well solution but at pH 4.0 and pH 7.0. Crystals were cryo-cooled at 100 K using as cryoprotectant synthetic buffer liquor with 8% ethylene glycol.

989

Honeybee PBP ph-Induced Domain Swapping

Structure determination and refinement

Acknowledgements

Diffraction data sets were collected at the laboratory with a MAR345dtb detector on a rotating anode (Bruker Microstar) and at the European Synchrotron Radiation Facility (Grenoble, France) (Table 2). The data were indexed and integrated using MOSFLM34 and scaled using SCALA.35 The crystals of the D35N mutant at pH 5.5 belong to space group C2221, with one molecule in the asymmetric unit. They are iso-structural to ASP1 with nBBSA [Protein Data Bank (PDB) code 3BJH],14 which was used as an initial model. The structures at pH 7.0 were solved by molecular replacement with MOLREP36 using the structure of ASP1 in complex with nBBSA at pH 5.5 (PDB code 3BJH) as the initial search model. Crystals belong to space group P212121 for the native protein and to space group P21212 for the mutant, with two molecules in the asymmetric unit. Structure refinement was performed with REFMAC5,36 alternating with rebuilding using COOT37 or Turbo-Frodo,38 using TLS parameters. In all cases, the electron density map of the ligands was clearly defined. Figures were made with PyMOL† or CCP4-MG.39 The binding site cavity volume was calculated using SURFNET.40

This work was supported in part by the PACA region, by the Région Ile de France (SESAME grant no. A01947) and by the Marseille-Nice-Genopole®. M.E. P. is a recipient of a joint PhD grant from the PACA region and the BioXtal Co. (no. ABDR/S05-003).

Fluorescence quenching experiments Fluorescence experiments were carried out on a Varian Eclipse spectrofluorimeter using a quartz cuvette in a right-angle configuration. The interactions of ASP1 with 9ODA and nBBSA were monitored by recording the quenching of the intrinsic protein fluorescence upon addition of ligand aliquots. The excitation wavelength was 290 nm, and emission spectra were recorded at 300– 400 nm. The excitation slit was 5 nm, while the emission slit was 10 nm. A moving-average smoothing procedure was applied, with a window of 3 nm. Titrations were carried out at 20 °C with 1 μM protein in 10 mM citrate disodium phosphate at pH 7.0 and pH 4.0. The fluorescence intensities at the maximum of emission (346 nm) for different concentrations of the quencher were corrected for the buffer contribution before plotting and further analysis. The affinity was estimated by plotting the decrease of fluorescence intensity at the emission maximum as 100 − (Ii − Imin)/(I0 − Imin) × 100 against the quencher concentration; I0 is the maximum of fluorescence intensity of the protein alone, Ii is the fluorescence intensity after the addition of quencher and Imin is the fluorescence intensity at the saturating concentration of quencher. The Kd values were estimated using Prism 4 (GraphPad Software, Inc.) by nonlinear regression for two-site binding with the equation, Y = Bmax1 × X/(Kd1 + X) + Bmax2 × X/ (Kd2 + X), where Bmax is the maximal binding and Kd is the concentration of ligand required to reach half-maximal binding. Accession codes The coordinates and reflection files have been deposited with the PDB at the Research Collaboratory for Structural Bioinformatics‡ with accession numbers 3CZ2, 3CYZ, 3CZ1, 3D73, 3D74, 3D75, 3D76, 3D77 and 3D78 (Table 2).

† http://pymol.sourceforge.net/ ‡ http://www.rcsb.org/pdb/

Supplementary Data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j. jmb.2009.05.067

References 1. Krieger, J., von Nickisch-Rosenegk, E., Mameli, M., Pelosi, P. & Breer, H. (1996). Binding proteins from the antennae of Bombyx mori. Insect Biochem. Mol. Biol. 26, 297–307. 2. Syed, Z., Ishida, Y., Taylor, K., Kimbrell, D. A. & Leal, W. S. (2006). Pheromone reception in fruit flies expressing a moth's odorant receptor. Proc. Natl Acad. Sci. USA, 103, 16538–16543. 3. Grosse-Wilde, E., Gohl, T., Bouche, E., Breer, H. & Krieger, J. (2007). Candidate pheromone receptors provide the basis for the response of distinct antennal neurons to pheromonal compounds. Eur. J. Neurosci. 25, 2364–2373. 4. Xu, P., Atkinson, R., Jones, D. N. & Smith, D. P. (2005). Drosophila OBP LUSH is required for activity of pheromone-sensitive neurons. Neuron, 45, 193–200. 5. Laughlin, J. D., Ha, T. S., Jones, D. N. & Smith, D. P. (2008). Activation of pheromone-sensitive neurons is mediated by conformational activation of pheromonebinding protein. Cell, 133, 1255–1265. 6. Vogt, R. G., Prestwich, G. D. & Riddiford, L. M. (1988). Sex pheromone receptor proteins. Visualization using a radiolabeled photoaffinity analog. J. Biol. Chem. 263, 3952–3959. 7. Jin, X., Ha, T. S. & Smith, D. P. (2008). SNMP is a signaling component required for pheromone sensitivity in Drosophila. Proc. Natl Acad. Sci. USA, 105, 10996–11001. 8. Benton, R., Vannice, K. S. & Vosshall, L. B. (2007). An essential role for a CD36-related receptor in pheromone detection in Drosophila. Nature, 450, 289–293. 9. Benton, R., Sachse, S., Michnick, S. W. & Vosshall, L. B. (2006). Atypical membrane topology and heteromeric function of Drosophila odorant receptors in vivo. PLoS Biol. 4, e20. 10. Sato, K., Pellegrino, M., Nakagawa, T., Vosshall, L. B. & Touhara, K. (2008). Insect olfactory receptors are heteromeric ligand-gated ion channels. Nature, 452, 1002–1006. 11. Wicher, D., Schafer, R., Bauernfeind, R., Stensmyr, M. C., Heller, R., Heinemann, S. H. & Hansson, B. S. (2008). Drosophila odorant receptors are both ligandgated and cyclic-nucleotide-activated cation channels. Nature, 452, 1007–1011. 12. Campanacci, V., Krieger, J., Bette, S., Sturgis, J. N., Lartigue, A., Cambillau, C. et al. (2001). Revisiting the specificity of Mamestra brassicae and Antheraea polyphemus pheromone-binding proteins with a fluorescence binding assay. J. Biol. Chem. 276, 20078–20084.

990 13. Lartigue, A., Gruez, A., Spinelli, S., Riviere, S., Brossut, R., Tegoni, M. & Cambillau, C. (2003). The crystal structure of a cockroach pheromone-binding protein suggests a new ligand binding and release mechanism. J. Biol. Chem. 278, 30213–30218. 14. Lartigue, A., Gruez, A., Briand, L., Blon, F., Bezirard, V., Walsh, M. et al. (2004). Sulfur single-wavelength anomalous diffraction crystal structure of a pheromonebinding protein from the honeybee Apis mellifera L. J. Biol. Chem. 279, 4459–4464. 15. Sandler, B. H., Nikonova, L., Leal, W. S. & Clardy, J. (2000). Sexual attraction in the silkworm moth: structure of the pheromone-binding protein–bombykol complex. Chem. Biol. 7, 143–151. 16. Lee, D., Damberger, F. F., Peng, G., Horst, R., Guntert, P., Nikonova, L. et al. (2002). NMR structure of the unliganded Bombyx mori pheromone-binding protein at physiological pH. FEBS Lett. 531, 314–318. 17. Damberger, F., Nikonova, L., Horst, R., Peng, G., Leal, W. S. & Wuthrich, K. (2000). NMR characterization of a pH-dependent equilibrium between two folded solution conformations of the pheromone-binding protein from Bombyx mori. Protein Sci. 9, 1038–1041. 18. Lautenschlager, C., Leal, W. S. & Clardy, J. (2005). Coil-to-helix transition and ligand release of Bombyx mori pheromone-binding protein. Biochem. Biophys. Res. Commun. 335, 1044–1050. 19. Kruse, S. W., Zhao, R., Smith, D. P. & Jones, D. N. (2003). Structure of a specific alcohol-binding site defined by the odorant binding protein LUSH from Drosophila melanogaster. Nat. Struct. Biol. 10, 694–700. 20. Thode, A. B., Kruse, S. W., Nix, J. C. & Jones, D. N. (2008). The role of multiple hydrogen-bonding groups in specific alcohol binding sites in proteins: insights from structural studies of LUSH. J. Mol. Biol. 376, 1360–1376. 21. Wogulis, M., Morgan, T., Ishida, Y., Leal, W. S. & Wilson, D. K. (2006). The crystal structure of an odorant binding protein from Anopheles gambiae: evidence for a common ligand release mechanism. Biochem. Biophys. Res. Commun. 339, 157–164. 22. Pesenti, M. E., Spinelli, S., Bezirard, V., Briand, L., Pernollet, J. C., Tegoni, M. & Cambillau, C. (2008). Structural basis of the honey bee PBP pheromone and pH-induced conformational change. J. Mol. Biol. 380, 158–169. 23. Zhou, J. J., Zhang, G. A., Huang, W., Birkett, M. A., Field, L. M., Pickett, J. A. & Pelosi, P. (2004). Revisiting the odorant-binding protein LUSH of Drosophila melanogaster: evidence for odour recognition and discrimination. FEBS Lett. 558, 23–26. 24. Kaissling (1985). Molecular basis of nerve activity. pp. 173–183, Walter de Gruyter, Berlin, Germany. 25. Du, G. & Prestwich, G. D. (1995). Protein structure encodes the ligand binding specificity in pheromone binding proteins. Biochemistry, 34, 8726–8732. 26. Leal, W. S., Chen, A. M., Ishida, Y., Chiang, V. P., Erickson, M. L., Morgan, T. I. & Tsuruda, J. M. (2005).

Honeybee PBP ph-Induced Domain Swapping

27.

28.

29.

30. 31.

32.

33.

34. 35.

36.

37. 38.

39.

40.

Kinetics and molecular properties of pheromone binding and release. Proc. Natl Acad. Sci. USA, 102, 5386–5391. Tegoni, M., Ramoni, R., Bignetti, E., Spinelli, S. & Cambillau, C. (1996). Domain swapping creates a third putative combining site in bovine odorant binding protein dimer. Nat. Struct. Biol. 3, 863–867. Horst, R., Damberger, F., Luginbuhl, P., Guntert, P., Peng, G., Nikonova, L. et al. (2001). NMR structure reveals intramolecular regulation mechanism for pheromone binding and release. Proc. Natl Acad. Sci. USA, 98, 14374–14379. Lautenschlager, C., Leal, W. S. & Clardy, J. (2007). Bombyx mori pheromone-binding protein binding nonpheromone ligands: implications for pheromone recognition. Structure, 15, 1148–1154. Grater, F., Xu, W., Leal, W. & Grubmuller, H. (2006). Pheromone discrimination by the pheromone-binding protein of Bombyx mori. Structure, 14, 1577–1586. Danty, E., Briand, L., Michard-Vanhee, C., Perez, V., Arnold, G., Gaudemer, O. et al. (1999). Cloning and expression of a queen pheromone-binding protein in the honeybee: an olfactory-specific, developmentally regulated protein. J. Neurosci. 19, 7468–7475. Vincentelli, R., Bignon, C., Gruez, A., Canaan, S., Sulzenbacher, G., Tegoni, M. et al. (2003). Mediumscale structural genomics: strategies for protein expression and crystallization. Acc. Chem. Res. 36, 165–172. Sulzenbacher, G., Gruez, A., Roig-Zamboni, V., Spinelli, S., Valencia, C., Pagot, F. et al. (2002). A medium-throughput crystallization approach. Acta Crystallogr., Sect. D: Biol. Crystallogr. 58, 2109–2115. Leslie, A. G. (1999). Integration of macromolecular diffraction data. Acta Crystallogr., Sect. D: Biol. Crystallogr. 55(Pt 10), 1696–1702. Collaborative Computational Project, Number 4. (1994). The CCP4 Suite: programs for crystallography. Acta Crystallogr., Sect. D: Biol. Crystallogr. 50, 760–763. Murshudov, G., Vagin, A. A. & Dodson, E. J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr., Sect. D: Biol. Crystallogr. 53, 240–255. Emsley, P. & Cowtan, K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 60, 2126–2132. Roussel, A. & Cambillau, C. (1991). The TURBOFRODO graphics package. Silicon graphics Geometry Partners Directory, vol. 81, pp. Silicon Graphics, Mountain View, CA. Potterton, L., McNicholas, S., Krissinel, E., Gruber, J., Cowtan, K., Emsley, P. et al. (2004). Developments in the CCP4 molecular-graphics project. Acta Crystallogr., Sect. D: Biol. Crystallogr. 60, 2288–2294. Laskowski, R. A. (1995). SURFNET: a program for visualizing molecular surfaces, cavities, and intermolecular interactions. J. Mol. Graphics, 13, 323–330.