Báo cáo khoa học: Cytokinin-induced structural adaptability of a Lupinus luteus PR-10 protein potx

Crystal and solution structures for several homologues have shown a similar overall fold with a vast internal cavity which, together with structural similarities to the steroidogenic acu

Lupinus luteus PR-10 protein

Humberto Fernandes1, Anna Bujacz2, Grzegorz Bujacz1,2, Filip Jelen3, Michal Jasinski1,

Piotr Kachlicki4, Jacek Otlewski3, Michal M Sikorski1and Mariusz Jaskolski1,5

1 Institute of Bioorganic Chemistry, Polish Academy of Sciences, Poznan, Poland

2 Institute of Technical Biochemistry, Faculty of Biotechnology and Food Sciences, Technical University of Lodz, Poland

3 Department of Protein Engineering, Faculty of Biotechnology, University of Wroclaw, Poland

4 Institute of Plant Genetics, Polish Academy of Sciences, Poznan, Poland

5 Department of Crystallography, Faculty of Chemistry, A Mickiewicz University, Poznan, Poland

On detection of various types of pathogens, such as

viruses, bacteria and fungi, or chemicals, such as

ethyl-ene or salicylic acid, which mimic the effect of pathogen

infection and thus induce stress [1], plants mount an

efficient defence programme in which a number of

genes are induced and expressed Among them are

genes coding the so-called pathogenesis-related (PR)

proteins [2], which have been grouped into 17 classes

according to their biological activity or physicochemical properties and sequence homology [2,3] PR proteins

do not constitute a superfamily of proteins, but rather represent a collection of unrelated protein families which function as part of the plant defence system [4] Most PR proteins are either secreted or localized in the vacuoles In contrast, PR proteins of class 10 (PR-10) are the only group that is intracellular and cytosolic [5]

Keywords

cytokinin; N,N¢-diphenylurea; plant

hormones; PR-10 proteins; yellow lupine

Correspondence

M Jaskolski, Department of

Crystallography, Faculty of Chemistry,

A Mickiewicz University, Grunwaldzka 6,

60–780 Poznan, Poland

Fax: +48 61 829 1505

Tel: +48 61 829 1274

E-mail: mariuszj@amu.edu.pl

(Received 3 December 2008, revised

8 January 2009, accepted 9 January

2009)

doi:10.1111/j.1742-4658.2009.06892.x

Plant pathogenesis-related (PR) proteins of class 10 are the only group among the 17 PR protein families that are intracellular and cytosolic Sequence conservation and the wide distribution of PR-10 proteins throughout the plant kingdom are an indication of an indispensable func-tion in plants, but their true biological role remains obscure Crystal and solution structures for several homologues have shown a similar overall fold with a vast internal cavity which, together with structural similarities

to the steroidogenic acute regulatory protein-related lipid transfer domain and cytokinin-specific binding proteins, strongly indicate a ligand-binding role for the PR-10 proteins This article describes the structure of a com-plex between a classic PR-10 protein [Lupinus luteus (yellow lupine) PR-10 protein of subclass 2, LlPR-10.2B] and N,N¢-diphenylurea, a synthetic cyto-kinin Synthetic cytokinins have been shown in various bioassays to exhibit activity similar to that of natural cytokinins The present 1.95 A˚ resolution crystallographic model reveals four N,N¢-diphenylurea molecules in the hydrophobic cavity of the protein and a degree of conformational changes accompanying ligand binding The structural adaptability of LlPR-10.2B and its ability to bind different cytokinins suggest that this protein, and perhaps other PR-10 proteins as well, can act as a reservoir of cytokinin molecules in the aqueous environment of a plant cell

Abbreviations

CPPU, N-phenyl-N¢-(2-chloro-4-pyridyl)urea; CSBP, cytokinin-specific binding protein; Hyp-1, phenolic oxidative coupling protein from Hypericum perforatum; ITC, isothermal titration calorimetry; LlPR-10.2B, Lupinus luteus (yellow lupine) PR-10 protein of subclass 2; N,N¢-DPU, N,N¢-diphenylurea; NCS, (S)-norcoclaurine synthase; PR, pathogenesis-related; StAR, steroidogenic acute regulatory protein; START, StAR-related lipid transfer.

PR-10 proteins, first identified in cultured parsley

cells [6], are small (155–163 residues), slightly acidic

and show resistance to proteases They are usually

encoded by multigene families (for instance, in yellow

lupine, there are 10 known genes encoding PR-10

pro-teins [7]), suggesting that a number of different protein

homologues can be expressed in various plant organs

under different conditions This makes the study of the

function of PR-10 proteins very complex

It is believed that PR-10 proteins are an essential

component of a plant defence programme because

their genes are usually induced by the attack of

vari-ous pathogens and by environmental stress Some

pr-10 genes are, however, expressed constitutively,

suggesting a more general biological role of PR-10

proteins in plant development [8–10] Several other

lines of evidence have implicated PR-10 proteins, or

at least proteins that are sequence-related to the

PR-10 class, in enzymatic functions, such as RNA

hydrolysis [11] and the synthesis of (S)-norcoclaurine

[12] or hypericin [13] However, it is obvious that

these functions are not universal among all PR-10

proteins Flores et al [14] have also postulated a

stor-age function for the PR-10 class Other functions

sometimes attributed to the PR-10 proteins include

antimicrobial activities, i.e antifungal [5,14–16],

anti-bacterial [14] and antiviral [15] activity Another

prop-osition postulates a hydrophobic ligand-binding role

for PR-10 class members

Crystal and solution structures are known for

sev-eral PR-10 homologues [17–25] They all reveal the

same general fold consisting of a seven-stranded

anti-parallel b-sheet wrapped around a long C-terminal

a-helix These two elements, together with two short

helices, form a hydrophobic cavity, the volume of

which is disproportionately large for proteins of this

size The abnormal size of the cavity has been taken

as an indication of an important role in hydrophobic

ligand binding Supporting evidence for such a

ligand-binding role is provided by the following

observations: (a) the structural similarity between

PR-10 members and the steroidogenic acute

regula-tory protein (StAR)-related lipid transfer (START)

domain of human MLN64 protein [26], which is a

steroid-binding domain related to StAR, involved in

cholesterol translocation in human placenta and

brain; (b) the crystal structure of Betv1 (a

birch-pollen PR-10 protein) complexed with deoxycholate

[21]; (c) the structural similarity with cytokinin-specific

binding protein (CSBP) [24]; and (d) the crystal

struc-ture of a yellow lupine (Lupinus luteus) homologue,

LlPR-10.2B, complexed with the adenine-type

cyto-kinin hormone trans-zeatin [25]

In this study, we have investigated whether the same LlPR-10.2B protein also has the ability to bind artifi-cial cytokinin molecules, chemically synthesized as urea derivatives Despite the apparent lack of chemical simi-larity between the adenine- and urea-type molecules, their cytokinin-type effect is very similar [27,28] The synthetic cytokinin investigated in the present work is N,N¢-diphenylurea (N,N¢-DPU) The crystal structure

of the LlPR-10.2B protein preincubated with N,N ¢-DPU shows that the protein is indeed able to store several N,N¢-DPU molecules in the internal cavity The protein’s ability to bind N,N¢-DPU and other cyt-okinins has also been studied by isothermal titration calorimetry (ITC) In addition, the antifungal activity

of the LlPR-10.2B protein has been investigated

Results

Asymmetric unit contents The crystal asymmetric unit contains one LlPR-10.2B protein molecule complexed with four N,N¢-DPU mol-ecules, 134 modelled water molecules and two sodium cations The metal cations have an octahedral coordi-nation close to loop L3 (Na1), where the ligands are three carbonyl O atoms (Pro31, Val34, Ile37) and three water molecules, and close to loop L9 (Na2), where the ligands are two carbonyl O atoms (Thr121, Gly123), the Oc1 atom of Thr121 and three water molecules The correctness of the interpretation of the metal sites as sodium is confirmed by the satisfactory refinement of the B factors (34.4 and 41.0 A˚2 for Na1 and Na2, respectively), by the final Na+ÆÆÆO distances (2.1–2.9 A˚) and by the bond-valence test [29]

Model quality and overall folding The refined 1.95 A˚ resolution crystallographic model

of the LlPR-10.2B protein has good overall geometry and Ramachandran statistics (Table 1) The quality of the electron density maps is high and there are only a few less clear areas in the loop regions and at the C-terminus The recombinant protein lacks the N-terminal methionine, excised during expression by Escherichia coli methionylaminopeptidase [30], whose catalytic efficiency is inversely proportional to the size

of the side-chain of the amino acid in the penultimate position (Gly in the LlPR-10.2B sequence)

The LlPR-10.2B molecule has an overall fold consisting, as in other PR-10 class proteins, of three a-helices (1–3) and seven antiparallel b-strands (1–7), giving rise to an a + b fold, whose most pronounced features are a b-grip over the C-terminal a3-helix and

a vast cavity bounded by all the helices and the

con-cave face of the b-sheet The strands of the b-sheet are

connected by b-hairpins, except for the connection

between the b1 and b2 edges of the sheet, which are

joined by a right-handed crossover involving the a1

and a2-helices There are seven b-bulges, which distort

the regularity of the b-sheet, endowing it with a highly

curved shape (Fig 1)

The C-terminal a3-helix is structurally and

sequen-tially the most divergent element of the PR-10

struc-ture (Figs 2 and 3A,B) This divergence is higher in

the N-terminal half of the helix The present protein,

LlPR-10.2B, is a close homologue of LlPR-10.2A,

characterized structurally by Pasternak et al [22]

Although the two sequences are nearly identical (91%

sequence identity), and there are no differences in the

N-terminal part of the a3-helix (Fig 2), the strong

inward collapse in the middle of the a3-helix, reported

for the ligand-free LlPR-10.2A molecule, is not

observed in the LlPR-10.2B protein complexed with cytokinins, including the present structure (Fig 3A,B) The straightening of the a3-helix significantly increases the volume of the internal cavity

One of the loops, L4, shows extraordinary rigidity despite the presence of four Gly residues in its sequence This Gly-rich loop is characterized by excel-lent electron density, low B factors and the highest degree of PR-10 sequence conservation (Fig 2) The sequence signature xEGxGGxGTx and structural conservation of the Gly-rich loop are observed even in distant homologues, such as CSBP The rigidity of the loop is maintained by a pattern of three hydrogen bonds between the Oc1 atom of Thr51 and the main-chain groups of Gly47 (N–H) and Gly48 (N–H and C=O)

The electron density maps for the four N,N¢-DPU ligands found in the internal cavity are of rather poor quality, contrasting with the excellent definition of the zeatin molecules in the LlPR-10.2B–zeatin complex [25] In the present complex, the N,N¢-DPU molecules were modelled in elongated ‘tubes’ of 2Fo) Fcelectron density, which did not have sufficient features for unequivocal assignment of the ligand atoms (Fig 4A) The evident disorder of the N,N¢-DPU molecules is the result of two factors: (a) the high degree of rotational freedom of the terminal phenyl rings (Fig 4B); and (b) the lack of specific interactions that would anchor the N,N¢-DPU molecules to the protein framework

Table 1 Data collection and refinement statistics.

Data collection

Cell parameters (A ˚ ) a = 34.4, b = 73.2, c = 100.0

Resolution limits (last shell) (A ˚ ) 40.0–1.95 (2.02–1.95)

No measured reflections 54070

Completeness (last shell) (%) 99.4 (96.8)

<I ⁄ r(I)> (last shell) 20.8 (3.8)

Refinement statistics

Resolution limits (A ˚ ) 15.0–1.95

No reflections in test set 860

No atoms

ÆBæ (A˚ 2

)

Rmsd from ideal

Chiral volumes (A ˚ 3 ) 0.106

Ramachandran u ⁄ w angles (%)

Fig 1 Overall fold of the LlPR-10.2B protein molecule with annota-tion of secondary structural elements The four N,N¢-DPU mole-cules found inside the binding cavity are shown in a ball-and-stick representation The binding cavity is represented as a mesh cast, calculated in VOIDOO [52] Sodium cations are represented as spheres This and all other structural illustrations have been prepared using PYMOL [53].

The only interatomic contacts of the N,N¢-DPU

ligands are three N–HÆÆÆO hydrogen bonds to water

molecules that are loosely positioned within the

bind-ing cavity (Fig 5) In view of the crude modellbind-ing, one

could question whether the inclusion of the four

N,N¢-DPU molecules is correct However, as the

crystalliza-tion buffer contained only N,N¢-DPU molecules and

citrate anions at high abundance, and as the ligand

electron density is definitely not compatible with the

branched structure of the hydrophilic citrate moiety,

the assumption that the binding cavity is occupied by

N,N¢-DPU molecules is the only logical conclusion

The binding cavity

The cavity enclosed within the LlPR-10.2B protein

core in the present N,N¢-DPU complex has a large

vol-ume, calculated as 3600 A˚3 by the surfnet program

[31], and can be accessed by two openings The larger

opening is located between the a3-helix and loops L3,

L5, L7 and L9, and is gated by a salt bridge between

Arg138 (a3-helix) and Glu59 (b3) and by a

water-bridged contact between the side-chains of Glu131

(a3-helix) and Thr93 (loop L7) The second opening is

found between the b1-strand and the a3-helix The

additional entrances seen near b5⁄ L7 ⁄ b6 and (a small one) near L2⁄ L4 in the crystal structure of the LlPR-10.2B–zeatin complex [25] are closed in the present N,N¢-DPU complex These differences are a result of a major conformational rearrangement in loop L7 (Fig 3C), and of smaller but significant changes in loops L2 and L4 Despite the closure of the b5⁄ L7 ⁄ b6 entrance in the present complex, a similar situation to that detected for the LlPR-10.2B–zeatin complex [25] exists, with a water molecule (W168 in the present complex) disrupting the b5–b6-sheet near loop L7 (Fig 6)

In addition to the different numbers of entrances, the cavities also have significantly different contents in the two complexes In the present LlPR-10.2B–N,N ¢-DPU complex, the cavity is occupied by four N,N ¢-DPU molecules (¢-DPU1–4) and eight water molecules (Fig 5), whereas three zeatin molecules and 25 water molecules fill the cavity of the LlPR-10.2B–zeatin com-plex In reality, there are less than four N,N¢-DPU molecules inside the cavity, as one of the phenyl rings

of the DPU4 molecule, pointing towards the b1–a3 opening (Fig 1), is outside the cavity A remarkable feature of N,N¢-DPU binding in the hydrophobic cavity is the lack of any specific interatomic

inter-Fig 2 Sequence alignment, calculated in CLUSTALW [54], for representative PR-10 proteins, for which crystal structures complexed with small-molecule ligands (or in apo form for LlPR10.2A) have been determined LlPR10.2B⁄ DPU, the present structure of a yellow lupine pro-tein complex with N,N¢-DPU; LlPR10.2B ⁄ zea, the same propro-tein complexed with trans-zeatin [25]; LlPR10.2A, a close homologue from yellow lupine [22]; VrCSBP⁄ zea, mung bean CSBP complexed with zeatin [24]; Betv1 ⁄ deox, birch pollen PR-10 protein complexed with deoxycho-late [21] Residues identical in all sequences are marked with an asterisk The symbols ‘:’ and ‘.’ below the sequences indicate conservative and semiconservative substitutions, respectively The differences between the two members of the yellow lupine PR-10.2 subfamily are marked with | The boxes outline regions that are most conserved (full line) or most divergent (broken line) The residues marked in grey interact with the ligands via van der Waals’ contacts, and those in black via hydrogen bonds The secondary structural elements correspond

to the LlPR10.2B ⁄ DPU structure Residues implicated in RNase activity in other PR-10 proteins are underlined in the LlPR10.2B sequence.

actions, such as hydrogen bonds, between the protein

and the N,N¢-DPU molecules (Figs 2 and 5)

The differences in stoichiometry as well as the

orien-tation and binding mode of the ligand molecules in the

two LlPR-10.2B complexes result in a rearrangement

of a number of side-chains that are pointing into the

interior of the cavity Clearly, some of these

rearrange-ments are simple translations following the shifts of

the Ca atoms that are connected with a remodelling of

the protein backbone (Phe5, Tyr19, Val23, Ile30,

Val34, Thr36, Ile37, Val40, Ile94, Ile97, Phe99,

Thr101, Val115, Pro127, Asn128) Other side-chains,

however, clearly have different conformations in the

two complexes, ranging from minimal changes for the

residues Asp7, Tyr9, Leu22, Phe57, Glu59, Lys64,

Val66, Tyr80, Tyr82, Leu103, Ile119 and Phe142, to

major conformational changes for the residues Lys53,

Ile84, Ile117, Glu131, Arg138 and Phe143 Another

interesting aspect is the number of residues, with

side-chains pointing into the cavity, that have been

modelled in double conformation Unlike the

LlPR-10.2B–zeatin complex, where only one residue (Val66)

has dual conformation, in the present complex two

cavity-forming residues (Leu55 and His68) are in double conformation

The LlPR-10.2B–N,N¢-DPU complex

In the present complex crystal structure, four N,N ¢-DPU molecules have been modelled in the electron density (Figs 1 and 4A) with full occupancies The average B factors of the N,N¢-DPU molecules are 31.8, 33.1, 34.2 and 45.0 A˚2 for DPU1, DPU2, DPU3 and DPU4, respectively

DPU1 is found deep in the cavity close to the a1 and a2-helices (Fig 1) It is anchored to the protein by several van der Waals’ contacts involving Leu22, Val23, Leu55, His68, Tyr80 and Phe142 DPU2 is placed near the a3⁄ L3 ⁄ L5 ⁄ L7 ⁄ L9 opening of the cavity (Fig 1) and establishes van der Waals’ contacts with Ile37, Phe57, Ile58 and Arg138 DPU3 is aligned with the a3-helix near the b1⁄ a3 opening (Fig 1) in an ori-entation that is roughly parallel to the DPU1 mole-cule It is anchored to the protein via van der Waals’ contacts with Tyr9, Leu22, Phe99, Thr101, Leu103, Gly113, Val115, Gly139 and Phe143 DPU3 also

α3

A

D

L9

L7

Fig 3 Superposition of PR-10 structures, calculated with Lsqkab [42] (A) Superposi-tion of four proteins with the PR-10 fold Colour code: yellow, LlPR-10.2B–N,N¢-DPU (Protein Data Bank code 3e85), this work; red, LlPR-10.2B–zeatin (2qim); magenta, LlPR-10.2A molecule A (1xdf); blue, CSBP– zeatin molecule B (2flh) The orientation emphasizes the differences in loops L7 and L9, and the C-terminal a3-helix (B)

‘Sausage’ representation of the deviations between the Ca atoms of LlPR-10.2B–N,N¢-DPU and LlPR-10.2A (C) ‘Sausage’ repre-sentation of the deviations between the Ca atoms of 10.2B–N,N¢-DPU and LlPR-10.2B–zeatin (D) Stereoview of a superposi-tion of LlPR-10.2B–N,N¢-DPU (yellow ⁄ green) and LlPR-10.2B–zeatin (red ⁄ aqua blue) com-plexes The C-terminal a3-helix has been omitted for clarity.

B

Fig 4 (A) The four N,N¢-DPU molecules present in the crystal structure of the LlPR-10.2B–N,N¢-DPU complex The 2F o ) F c electron density maps are contoured at the 1.0r level (B) Conformation of the N,N¢-DPU molecules shown in a superposition of the urea moiety The atom numbering scheme of the N,N¢-DPU molecule is also shown.

Fig 5 Illustration of the four N,N¢-DPU

molecules found in the protein cavity,

adapted from a LIGPLOT representation [55].

Hydrogen bonds are indicated by broken

lines and van der Waals’ contacts by radial

lines Water molecules are shown as small

grey spheres.

makes an indirect hydrogen bond contact with the

pro-tein (Tyr82), mediated by a water molecule (W197)

(Fig 5) DPU4 is placed at the N-terminal part of the

a3-helix, with the two phenyl rings pointing at the two

different entrances to the cavity, namely b1⁄ a3 and

a3⁄ L3 ⁄ L5 ⁄ L7 ⁄ L9 (Fig 1) It is anchored to the protein

via van der Waals’ contacts with Phe5, Gly89, Phe99,

Ile117, Gly132, Ala135 and Arg138 The N,N¢-DPU

ligands also establish van der Waals’ contacts between

themselves, namely within the pairs DPU1–DPU2 and

DPU1–DPU3

The N,N¢-DPU molecules have a rigid and planar

carbonyl diamide group with two lateral phenyl

substi-tutes at different orientations (Fig 4B) In their

elon-gated electron density, the N,N¢-DPU molecules can

move along their long axes This is consistent with the

lack of direct hydrogen bond interactions with the

protein scaffold

The N- and C-termini

In the PR-10 protein topology, the amino terminus

forms the b1-strand of the b-sheet and is typically

highly ordered In the present LlPR-10.2B–N,N¢-DPU

complex, the presence of the sodium cation (Na2)

coordinated in loop L9 pushes the first two amino acid

residues of b1 away from the b-sheet Because of the

presence of the cation, the N-terminal NH3+group is

not stabilized by the conserved hydrogen bonds to

resi-dues in loop L9 [(Thr121)O, (Gly123)O, (Thr121)Oc1]

seen in other yellow lupine PR-10 structures [20,22,25]

Nonetheless, the N-terminal part of the protein is not

disordered as it has a very clear electron density The

presence of Na2 also induces a shift of loop L9 relative

to the LlPR-10.2B–zeatin structure, where no cation

was observed in this region (Fig 3A,C) The sodium

cation Na2 is thus responsible for conformational changes at both loop L9 and the N-terminus of the protein However, the carboxyl terminus is less ordered, and no electron density is visible for the last three residues Consequently, Pro154 is the last amino acid included in the model

Structural comparisons of PR-10 proteins All the known models of PR-10 proteins have a com-mon canonical fold, consisting of a seven-stranded antiparallel b-sheet wrapped around a long C-terminal a-helix These elements, together with two short heli-ces, form a large hydrophobic cavity Because of its unusual volume, particularly in view of the small size

of the PR-10 proteins, the large cavity is assumed to have evolved for hydrophobic ligand binding

Despite the same overall fold, the superposition of the different PR-10 models reveals interesting struc-tural differences The most significant is the strong inward kink of the C-terminal a3-helix observed for the ligand-free LlPR-10.2A protein [22], but not in the trans-zeatin complex of LlPR-10.2B [25] As the sequences of the LlPR-10.2A and LlPR-10.2B homo-logues are 96% similar, and there are no differences in the sequence surrounding the point of bending (Phe142) (Fig 2), the observed structural difference was assumed to be a manifestation of the adaptability

of the PR-10 fold to the presence of the hormone ligands [25] As a consequence of the straightening of the a3-helix, the cavity of LlPR-10.2B complexed with trans-zeatin has a volume of 4500 A˚3, in contrast with the volume of only 2000 A˚3observed in the ligand-free LlPR-10.2A structure This ligand-induced adaptability hypothesis is reinforced by the present crystal structure

of the LlPR-10.2B–N,N¢-DPU complex, which shows

Fig 6 Disruption of the antiparallel b-sheet

in the region of the b5–b6-strands by a water molecule (A) Ball-and-stick represen-tation and 2F o ) F c electron density map contoured at the 1.3r level (B) Cartoon representation of the protein with the region shown in (A) highlighted in dark blue.

the same difference, although of a smaller magnitude

(Fig 3A,C) The volume of the cavity in the present

LlPR-10.2B–N,N¢-DPU complex is 3600 A˚3, i.e it is

smaller than in the zeatin complex, but still much

larger then in apo LlPR-10.2A These observations

indicate an induced structural flexibility of yellow

lupine PR-10 proteins, and perhaps also of

homo-logues from other species Comparison of all the

struc-tures of PR-10 proteins demonstrates a huge

variability in the volume of the cavity, which ranges

from 4500 A˚3 in the LlPR-10.2B–zeatin complex to

1100 A˚3in the CSBP–zeatin complex (molecule C)

In addition to the difference in the cavity volume,

the two complexes of the LlPR-10.2B protein, one with

N,N¢-DPU and the other with trans-zeatin, show a

dif-ference in the identity and coordination sites of metal

ions In contrast with the two sodium cations

coordi-nated close to loops L3 and L9 in the present

LlPR-10.2B–N,N¢-DPU complex, in the LlPR-10.2B–zeatin

complex only one metal site close to loop L3, and

identified as calcium, was detected [25] Metal cations

were found in two other PR-10 protein structures,

namely in LlPR-10.2A [22] and CSBP [24] (sodium in

both cases), where they were coordinated by the L3

and L9 loops, respectively

Cytokinin-binding assays

To verify the structurally derived data, we used direct

measurement of the thermodynamic parameters of

LlPR-10.2B–ligand interactions The cytokinin-binding

capability of LlPR-10.2B was tested by ITC for

natu-ral adenine-type (trans-zeatin, kinetin) and artificial

urea-type [N,N¢-DPU,

N-phenyl-N¢-(2-chloro-4-pyr-idyl)urea (CPPU)] hormones The ITC assays were

performed using different protein concentrations and

different experimental conditions, as summarized in

Experimental procedures Under the experimental

con-ditions, limited by the low solubility of the ligands, the

ITC data show that the LlPR-10.2B protein can bind

all four cytokinins The data from the calorimetric

measurements were analysed by fitting

independent-and multiple-site models, as well as a cooperative

model, to the raw data Only the independent-site

model provided meaningful results The LlPR-10.2B

protein has one binding site characterized by a Kd

value in the micromolar range and a different

stoichi-ometry depending on the ligand molecule

LlPR-10.2B–N,N¢-DPU binding is exothermic with a 1 : 1

stoichiometry (n = 0.83) and a Kd value of 4.2 ±

1.6 lm (Fig 7) Titration with trans-zeatin revealed an

additional binding site, with a Kd value in the

submil-limolar range Thus, the LlPR-10.2B–zeatin complex is

characterized by two binding sites, one with a Kdvalue

of 12.3 ± 5.1 lm and 1 : 1 stoichiometry, and the other with a Kdvalue of 193 ± 43.7 lm and 8 : 1 stoi-chiometry CPPU and kinetin show clear interaction with the LlPR-10.2B protein; however, the thermo-dynamic parameters could not be determined accu-rately because of the low solubility of both ligands

Antifungal assays Many PR proteins have long been known to have anti-fungal or antibacterial activities [32–34], but, until

Fig 7 Calorimetric titration of LlPR-10.2B with N,N¢-DPU The top panel shows raw heat data corrected for baseline drift obtained from 23 consecutive injections of 3.02 m M N,N¢-DPU into the sample cell (750 lL) containing 0.37 m M LlPR-10.2B protein in

3 m M citrate buffer, pH 6.3, at 20 C The bottom panel shows the binding isotherm created by plotting the heat peak areas against the molar ratio of N,N¢-DPU added to LlPR-10.2B present in the sample cell The heats of mixing (dilution) were subtracted The line represents the best fit to a model of n independent sites LlPR-10.2B–N,N¢-DPU binding is exothermic with 1 : 1 stoichiometry (n = 0.83) and a Kdvalue of 4.2 ± 1.6 l M

recently, the PR-10 family was not included in this

group The antifungal activity of PR-10 proteins was

first demonstrated in 2002 for ocatin, a PR-10

homo-logue from the Andean tuber crop oca (Oxalis tuberose

Mol.) [14], later for the hot pepper (Capsicum annuum)

CaPR-10 protein [15], and recently for the yellow-fruit

nightshade (Solanum surattense) SsPR-10 protein [16]

and the peanut (Arachis hypogaea) AhPR-10 protein

[5] In general, recombinant proteins were tested for

their ability to inhibit specific fungal growth

The effect of recombinant LlPR-10.2B on the

in vitro growth of pathogenic fungi has been

investi-gated in this work For the assays, a lupine-specific

fungus and two fungi specific for other plants

belong-ing to the same class (Magnoliopsida) were used,

namely Colletotrichum lupini, Leptosphaeria maculans

and Leptosphaeria biglobosa Purified recombinant

LlPR-10.2B protein did not inhibit the growth of any

of the fungi under study (not shown)

Discussion

The search for the biological role of the PR-10

pro-teins has focused recently on the abnormal size of the

internal cavity, which could function as a binding

site⁄ reservoir for hydrophobic ligands in the aqueous

environment of the plant cell This proposition finds

support in the structural similarity between classic

PR-10 proteins and ligand-binding proteins such as

CSBP and the START domain [24,26] In addition,

several studies have demonstrated the ability of PR-10

proteins to bind steroids, cytokinins, fatty acids and

flavonoids [21,35,36] A different possible function of

PR-10 proteins has emerged from several other studies,

connected with the enzymatic biosynthesis of

second-ary metabolites, such as (S)-norcoclaurine [12] or

hypericin [13] The enzymes catalysing the above

reac-tions [(S)-norcoclaurine synthase (NCS) and phenolic

oxidative coupling protein from Hypericum perforatum

(Hyp-1), respectively] have been postulated to belong

to the PR-10 structural class of proteins based only on

sequence comparisons (sequence identity about 38%

for NCS and 45% for Hyp-1) Recently, these

predic-tions have been confirmed for NCS [37], but further

work will be necessary to verify these assumptions for

the Hyp-1 protein [38] However, in view of their high

level of expression, a universal catalytic function for

all PR-10 proteins seems unlikely

The present work provides structural evidence that

classic PR-10 proteins can bind N,N¢-DPU, the first

synthetic cytokinin to be identified [27] Cytokinins are

structurally diverse and biologically versatile

phytohor-mones, involved in the differentiation of shoot

meri-stem and root tissues, leaf formation and senescence, chloroplast development, etc [39] The natural cytoki-nins are adenine derivatives and can be classified by the character of their N6 substituent as isoprenoid, aromatic or furfuryl cytokinins Cytokinins with an unsaturated isoprenoid side-chain are by far the most prevalent, in particular those with a trans-hydroxylated substituent (trans-zeatin and its derivatives) Various phenylurea derivatives constitute a group of synthetic cytokinins, some of which are highly active, e.g N,N ¢-DPU, CPPU or thidiazuron [28] The phenylurea derivatives have been shown to exhibit biological activ-ity very similar to that of N6-substituted adenine derivatives in various cytokinin bioassays These com-pounds were developed for commercial use as defoli-ants in cotton and other crops, and are now widely used as cytokinins in higher plant tissue cultures and micropropagation protocols [40] The similarity of the biological activity of two structurally unrelated classes

of compounds has posed one of the more interesting problems in the study of cytokinin structure–function relationships [41]

The present LlPR-10.2B–N,N¢-DPU complex is the second example of a cytokinin complex of this protein Recently, we reported the crystal structure of LlPR-10.2B complexed with trans-zeatin [25] The binding mode of the two cytokinins is different (Figs 2 and 3D) which, in view of the chemical difference between these ligands, is not surprising The largest difference

is in the stoichiometry of the complexes In the present LlPR-10.2B–N,N¢-DPU complex, four N,N¢-DPU mol-ecules are accommodated in the hydrophobic cavity of the protein, in contrast with the 3 : 1 stoichiometry of the complex with trans-zeatin This remarkable ability

of the hydrophobic cavity of PR-10 proteins to hold a variable number of ligand molecules is highlighted when the CSBP–zeatin complex is considered, in which

a single CSBP molecule can bind either one or two trans-zeatin ligands [24]

The second aspect that differentiates the two LlPR-10.2B complexes is the protein–ligand interactions Although, in the case of trans-zeatin, several hydrogen bonds anchor the ligand molecules to the protein [25] (Fig 2), the N,N¢-DPU molecules interact with the protein virtually exclusively via van der Waals’ con-tacts (Figs 2 and 5) Thirdly, there is no direct corre-spondence between the ligand molecules in the two complexes In general terms, the N,N¢-DPU molecules 1–3 and zeatin 1–3 occupy similar spatial positions inside the protein cavity, but without individual over-lap (Fig 3D) DPU4 is placed at a distinct position that does not overlap with any zeatin molecule (Fig 3D)

A structural superposition of the LlPR-10.2B

mole-cules in the two complexes reveals that their Ca traces

show small but significant differences, despite the same

overall fold (Fig 3A) The rmsd between their Ca

coordinates is 0.77 A˚, with the major differences

local-ized in loop L7 (maximum deviation of 9.3 A˚ for the

Ca atoms of Gly89) and loop L9 (maximum deviation

of 4.7 A˚ at Gly123) The conformational change of

loop L9 is evidently caused by the presence of the

sodium cation (Na2) coordinated in this area in the

LlPR-10.2B–N,N¢-DPU structure In the LlPR-10.2B–

N,N¢-DPU complex, this cation disrupts the stabilizing

hydrogen bond between the N-terminus and loop L9

that is observed in all other yellow lupine PR-10

struc-tures [20,22,25], as well as the typical b-sheet

associa-tion with the b7-strand This disrupassocia-tion, however, does

not create any major disorder of the N-terminus as the

residues have excellent definition in the electron density

map The other sodium cation (Na1) is coordinated by

loop L3 at the same position as the calcium ion in the

LlPR-10.2B–zeatin structure [25] The change in metal

identity is most probably caused by the high sodium

concentration in the crystallization buffer The sodium

site at loop L9 is new in this yellow lupine PR-10

structure, but it is interesting to note that the same site

was occupied by a metal cation in the crystal structure

of mung bean CSBP [24]

As described above, the major folding differences

between the two LlPR-10.2B models are localized at

loop L7 This reshaping cannot be explained by the

different packing modes (C2221 and P65 for

LlPR-10.2B–N,N¢-DPU and LlPR-10.2B–zeatin,

respec-tively), and thus it must be concluded that it results

from the different ligand cargo present in the cavity

In both cases, the lattice interactions are weak, with

only one salt bridge (Asp92ÆÆÆLys20) present in the

LlPR-10.2B–N,N¢-DPU structure, and one main-chain

hydrogen bond (Gly89ÆÆÆVal2) in the LlPR-10.2B–

zeatin structure

More revealing than a simple Ca alignment in this

case of identical ligand-binding structures is an

all-atom superposition of the protein scaffolds Such a

superposition, calculated in Lsqkab [42], is

character-ized by an rmsd of 1.8 A˚ The largest difference of

13.5 A˚ is found for the Cc2 atoms of Leu90

Several reports have shown recently that PR-10

pro-teins can exhibit antifungal properties [5,14–16] Some

authors have associated the antifungal activity of

PR-10 proteins with their purported RNase activity

Chadha and Das [5] reported, for example, that these

activities are linked in AhPR-10, a PR-10 protein from

peanut Previously, Moiseyev et al [43] hypothesized

that the residues Lys54, Glu96, Glu148 and Tyr150

(ginseng ribonuclease 1 sequence) are responsible for RNase activity, and thus can be expected to be vital for antifungal activity As the indicated residues are fully conserved in the LlPR-10.2B protein sequence (Fig 2), antifungal activity could be expected for this protein as well However, in our experiments, no anti-fungal activity could be observed for any of the fungi tested (C lupini, L maculans and L biglobosa) The fungus C lupini is specific for lupine plants and infects the leaves The remaining two fungi are pathogens of oilseed rape (Brassica napus)

The two crystal structures of LlPR-10.2B show that

it can serve as a ligand-binding protein The structures provide evidence of the ability of the LlPR-10.2B pro-tein to bind cytokinins, either natural (trans-zeatin) or synthetic (N,N¢-DPU) Under the experimental condi-tions, limited by the low solubility of the ligands in water-based buffers, ITC binding affinity data for the four cytokinins were obtained Two of them represent the synthetic group (N,N¢-DPU and CPPU) and the other two are natural cytokinins, with one example of isopentenyl (trans-zeatin) and one of furfuryl (kinetin) N6-substituted adenine

The three crystal structures of PR-10-type proteins complexed with cytokinins, namely CSBP–zeatin [24], LlPR-10.2B–zeatin [25] and LlPR-10.2B–N,N¢-DPU (present work), together with cytokinin-binding affinity studies for the CSBP [24] and LlPR-10.2B (present work) proteins, provide a broad view of the inter-actions of these plant hormones with PR-10 and PR-10-like folded proteins One of the binding sites for trans-zeatin observed for the LlPR-10.2B protein has a similar Kd value (193 lm) to that obtained for CSBP (106 lm) [24] However, the ITC-determined stoichi-ometry is different The other binding site observed for LlPR-10.2B with all four cytokinin ligands, character-ized by a Kd value in the micromolar range, was not observed for CSBP The CSBP–N,N¢-DPU complex has not been investigated, and so no comparison with LlPR-10.2B can be made However, both proteins interact with kinetin and CPPU, suggesting that a CSBP–N,N¢-DPU complex is also possible

The different ligand-binding characteristics of the LlPR-10.2B and CSBP proteins may be a result of the large difference in the volume of the binding cavity (3600–4500 A˚3 for LlPR-10.2B, 1100–1600 A˚3 for CSBP), as measured by the surfnet program [31] The smaller cavity volume of CSBP results from the C-terminal a3-helix being less separated from the b-grip, and the a1 and a2-helices being closer to the centre of the protein

The crystal structure of the LlPR-10.2B–N,N¢-DPU complex shows four N,N¢-DPU molecules inside the