Báo cáo Y học: Interactions between the Fyn SH3-domain and adaptor protein Cbp/PAG derived ligands, effects on kinase activity and affinity docx

PTKs of the Src kinase family employ a well-conserved modular arrangement of interaction domains in the regulation of kinase Keywords kinase activity; proline-rich motifs; protein– prote

protein Cbp/PAG derived ligands, effects on kinase

activity and affinity

Silje A Solheim1,2, Evangelia Petsalaki3, Anne J Stokka1,2, Robert B Russell3, Kjetil Taske´n1,2 and Torunn Berge1,2

1 The Biotechnology Centre of Oslo, Norway

2 Centre for Molecular Medicine Norway, Nordic EMBL Partnership, University of Oslo, Norway

3 European Molecular Biology Laboratory, Heidelberg, Germany

Tyrosine phosphorylation is one of the key regulatory

protein modifications in multicellular organisms and

tightly controls and coordinates a wide range of

cellu-lar responses such as growth, metabolism, tissue

repair, migration and apoptosis [1–3] Phosphorylation

modulates enzymatic activity as well as creating new

binding sites for the recruitment of active molecules

into signalling complexes, and assists in building

dynamic networks for the transduction of information

from the extracellular environment to intracellular

signalling pathways Accurate and specific processing

of information is vital to maintaining cellular homeo-stasis, and errors in signal transduction pathways are linked to a range of diseases such as cancer, auto-immunity and diabetes [4,5]

Tyrosine phosphorylation is a reversible modifica-tion, regulated by protein tyrosine kinases (PTKs) and phosphatases (PTPs) [6,7] PTKs of the Src kinase family employ a well-conserved modular arrangement

of interaction domains in the regulation of kinase

Keywords

kinase activity; proline-rich motifs; protein–

protein interactions; SH3 domain; tyrosine

phosphorylation

Correspondence

T Berge, The Biotechnology Centre of Oslo,

Gaustadalleen 21, N-0319 Oslo, Norway

Fax: +47 2284 0501

Tel: +47 2284 0519

E-mail: torunn.berge@biotek.uio.no

Website: http://www.biotek.uio.no

(Received 8 May 2008, revised 24 June

2008, accepted 4 August 2008)

doi:10.1111/j.1742-4658.2008.06626.x

Csk-binding protein⁄ phosphoprotein associated with glycosphingolipid-enriched domains is a transmembrane adaptor protein primarily involved

in negative regulation of T-cell activation by recruitment of C-terminal Src kinase (Csk), a protein tyrosine kinase which represses Src kinase activity through C-terminal phosphorylation Recruitment of Csk occurs via SH2-domain binding to PAG pTyr317, thus, the interaction is highly dependent

on phosphorylation performed by the Src family kinase Fyn, which docks onto PAG using a dual-domain binding mode involving both SH3- and SH2-domains of Fyn In this study, we investigated Fyn SH3-domain bind-ing to 14-mer peptide ligands derived from Cbp⁄ PAG-enriched micro-domains sequence using biochemical, biophysical and computational techniques Interaction kinetics and dissociation constants for the various ligands were determined by SPR The local structural impact of ligand association has been evaluated using CD, and molecular modelling has been employed to investigate details of the interactions We show that data from these investigations correlate with functional effects of ligand binding, assessed experimentally by kinase assays using full-length PAG proteins as substrates The presented data demonstrate a potential method for modula-tion of Src family kinase tyrosine phosphorylamodula-tion through minor changes

of the substrate SH3-interacting motif

Abbreviations

Cbp ⁄ PAG, Csk binding protein ⁄ phosphoprotein associated with glycosphingolipids-enriched microdomains; Csk, C-terminal Src kinase; GST, glutathione S-transferase; PPII, polyproline type II; PRD, proline-rich domain; PTK, protein tyrosine kinase; SFK, Src family kinase; SH2, Src homology 2; SH3, Src homology 3; TCR, T-cell receptor.

activity and inhibition, as well as in the combinatorial

assembly of active signalling complexes The common

structure consists of an N-terminal

membrane-target-ing region (SH4), a Src-homology 3 (SH3) domain and

a Src-homology 2 (SH2) domain, capable of binding

to proline-rich motifs and phosphotyrosine residues,

respectively These interaction domains are followed

successively by a tyrosine kinase (SH1) domain [8,9]

In addition, Src family kinases (SFKs) contain both a

C-terminal auto-inhibitory phosphorylation site and an

activating auto-phosphorylation site in the kinase

domain Interactions occur frequently via individual

domains; however, both domains may also cooperate

to facilitate specific and stable complex formation [10–

12]

In T cells, the SFKs Lck and Fyn have central roles

in early signal transduction events as the most

proxi-mal signalling molecules to be activated downstream

of the T-cell receptor following direct interaction

between the receptor and peptide–MHC complexes on

antigen presenting cells This activation leads to a

cas-cade of tyrosine phosphorylation-dependent signalling

pathways [2,3,13] Fyn associates with and

phosphory-lates the transmembrane adaptor protein Csk-binding

protein⁄ phosphoprotein associated with

glycosphingo-lipid-enriched microdomains (Cbp⁄ PAG) [14], which is

localized to lipid rafts by palmitoylation anchoring

[15] The adaptor functions primarily as a negative

reg-ulator of SFKs and T-cell activation via recruitment of

C-terminal Src kinase (Csk) to PAG pTyr317 (human)

[16–18], however, the regulatory role of PAG may be

more complex as the adaptor may also act as an

acti-vating partner of Lyn, an SFK, and pSTAT3 in B-cell

lymphomas, as suggested by Tauzin et al [19]

The involvement of either SH3 or SH2 domains has

been discussed for the association of Fyn with PAG

[20,21]; in fact, Fyn utilizes both domains in this inter-action, as we have demonstrated previously [22] Bind-ing of the Fyn SH3-domain to the first proline-rich region of PAG is essential for the initiation of PAG tyrosine phosphorylation, which occurs via a proces-sive phosphorylation mechanism [23] This sub-sequently allows binding of the Fyn SH2-domain to PAG Tyr163 or pTyr181 and results in a dual-domain docking, enhancing the affinity of the Fyn–PAG inter-action and rendering Fyn insensitive to negative regu-lation by Csk (Fig 1A)

In this study, we explore further the initial step of the Fyn–PAG association; binding of the kinase SH3-domain to the first proline-rich region of PAG Full-length PAG variants, developed using a 2D peptide array approach, have previously been employed to demonstrate the functional effects of the Fyn–PAG complex in T cells [22] Using full-length proteins as templates, we created 14-mer peptide ligands (Fig 1B), one corresponding to the wild-type PAG interaction region (PRD1), and two structural variants of this sequence with low RLP*) and high (PRD1-super) affinity for the Fyn SH3-domain, respectively PAG also has a second proline-rich region, known to interact with the SH3-domain of the related kinase Lyn [24], and a 14-mer peptide (PRD2) containing this binding site was included as a control We conducted binding assays of peptide ligands using the isolated Fyn SH3-domain (Fig 1C) as well as full-length Fyn affinity precipitated from human primary T-cell lysates Kinetics and specificity of Fyn SH3-domain interactions were determined using SPR, and CD spec-troscopy was used to assess local structural changes upon SH3-domain binding The tertiary structure of the Fyn SH3-domain complexed to ligand is known both from X-ray diffraction and NMR studies [25,26],

A

C

B

Fig 1 The Fyn-PAG interaction (A) Schematic representation showing Fyn binding to the transmembrane adaptor PAG via SH3- and SH2-domain interactions (B) The 14-mer peptide ligands used in this study include the first (PRD1) and second (PRD2) proline-rich regions

of PAG as well as variants of the first region (PRD1-RLP* and PRD1-super) developed using 2D peptide arrays [22] (C) Secondary structure elements of the Fyn SH3-domain The five b strands, numbered consecutively, are separated by the interaction loops (bold) making primary contacts with the ligand Trp119 (*) is a highly conserved residue among SFK SH3-domains.

and this information was used to model the

pep-tide⁄ SH3-domain complexes in silico to investigate

details of the ligand interactions We show that

changes in affinity and secondary structure have

signif-icance for the functional effects of ligand binding,

assessed experimentally by kinase assays using

full-length PAG proteins as substrates

Results and Discussion

Phosphorylation of the adaptor protein PAG is

depen-dent on an initial interaction between the first

proline-rich region of PAG and the SH3-domain of the kinase

Fyn Our earlier investigations have highlighted the

importance of the Fyn SH3-domain interaction in

initi-ating the association with PAG, and how this affects

tyrosine phosphorylation and thereby the functionality

of PAG as a negative regulator of T-cell activation

through recruitment of Csk, dependent on PAG

pTyr317 [22] In this study, we recreated this

associa-tion in vitro using isolated Fyn SH3-domains and

syn-thetic peptide ligands containing wild-type PAG

sequences as well as structural variants developed

using a 2D peptide array approach, a rapid and

semi-quantitative approach for evaluation of amino acid

substitutions of all residues in a defined region

The Fyn–PAG interaction: association via the Fyn

SH3-domain

Several studies have focused on the identification of

high-affinity ligands for SH3-domains, either through

phage display or combinatorial peptide libraries [27–

29] Combined with data from NMR and X-ray

crys-tallography [25,26,30,31], these reports have revealed

in detail many of the requirements of SH3-domain

ligand binding, and optimal core recognition motifs

have been established for several kinase SH3-domains,

such as RPLPPLP for Src, Fyn, Lyn and

phosphati-dylinositol 3-kinase This motif is a class I SH3

consensus motifs (RxxPxxP), characterized by an

N-terminal arginine residue known to form an

orienta-tion-determining salt bridge with a key aspartate

resi-due [27,28,32] Interestingly, the Fyn SH3 domain

contains two such adjacent residues Asp99 and

Asp100, which may both participate in formation of

this salt bridge [33] Using 2D peptide array analyses,

the minimal interaction sequence of the wild-type

PAG-derived peptide PRD1 has been established to

RELPRIP [22], i.e four residues in common with the

phage display motif, whereas the high-affinity

PRD1-superpeptide (Glu132Pro, Pro135Arg of PAG, human)

of this study matched the minimal motif with five

common residues By contrast, the low-affinity sequence PRD1-RLP*, lacked critical amino acids by three alanine substitutions of the minimal motif (Arg131, Leu133 and Pro134 of PAG, human), and PRD2 (PPPVPVK), the peptide corresponding to the second proline-rich region of PAG, had only two residues in common with the minimal core motif

To analyse further the implications of PAG binding

to the Fyn SH3-domain, 14-mer peptides were synthes-ised consisting of these sequence variants of the inter-action site (PRD1, PRD1-RLP* and PRD1-super) as well as PRD2 (Fig 1B) Interaction data obtained using full-length proteins were first verified by per-forming biotin–streptavidin affinity precipitations on human T-cell lysates using biotin-tagged peptide ligands (Fig 2) The pull-down assays demonstrated a significant reduction in binding using the PRD1-RLP* peptide, whereas the PRD1-super peptide, developed as

a high-affinity Fyn SH3 binder using the 2D peptide array approach, displayed an  2.5-fold increase in Fyn binding relative to the wild-type (PRD1) peptide Interaction with the first proline-rich sequence was exclusive as no association was observed in pull-down assays performed with the peptide PRD2 containing the second PAG proline-rich domain

Conformation of peptide ligands in solution Polyproline sequences are known to adopt a distinct secondary structure in aqueous solutions due to the unique conformational properties of the cyclic imino nature of proline residue [34,35] This structure, a polyproline type II helix (PPII), has a specific geo-metry exploited in SH3-domain ligand binding that enables direct interactions with the hydrophobic bind-ing grooves of this domain Usbind-ing CD spectroscopy,

we investigated the extent of PPII helix conformation

of the peptide ligands PRD1, PRD1-RLP*,

Fig 2 Fyn interacts with the first proline rich region in PAG Lysates from human primary T cells were incubated with the indi-cated biotin-linked peptides and complexes precipitated using excess streptavidin beads Interactions were evaluated by western blot analysis using anti-Fyn IgG.

PRD1-super and PRD2 (Fig 3) In water, typical CD

spectra of the PPII helix show a negative band at 200–

205 nm and a positive band at 217–225 nm [36]

Although the positive band was absent from our

sam-ples, all four peptides displayed strong negative bands

at  201 nm, indicating that they are not completely

unstructured in aqueous buffer The negative maxima,

believed to arise from the extent of PPII conformation,

was found to correspond with the number of proline

residues in each peptide sequence (PRD2 >

PRD1 = PRD1-super > PRD1-RLP*) The absence

of a characteristic positive shoulder from all spectra

was thought to arise from the presence of other

sec-ondary structure elements in the samples [34,35]

Effects of PAG–Fyn SH3 interaction modulation

The functional effects of modulating the affinity of

the PAG–Fyn SH3 interaction were examined in an

in vitro kinase assay using full-length, recombinant

PAG protein and active Fyn kinase In T cells, PAG

has a dual role as both a ligand and a substrate for

Fyn, and with nine potential tyrosine phosphorylation

sites the adaptor is an excellent subject for quantitative

analysis of phosphorylation effects mediated via

SH3-domain interactions

In a pulse-chase assay, active Fyn and recombinant

PAG were incubated on ice with a low concentration

of radioactive ATP After 10 min, a large excess of

ATP was added and the phosphorylation process was

continued with aliquots withdrawn at the indicated

time points as illustrated for wild-type PAG in

Fig 4A Phosphorylated PAG species appeared as

dis-tinct bands, with the slower migrating species

corre-sponding to more heavily phosphorylated protein

accumulating over time Quantitative analyses of PAG phosphorylation of the three variants were performed over a time course of 5 min by measuring pixel inten-sity of the highest band at each time point relative to fully phosphorylated wild-type PAG (Fig 4B) The wild-type PAG protein was rapidly phosphorylated by Fyn, reaching the highest molecular mass species after

 1 min The full-length high affinity SH3-binder, superPAG, was phosphorylated in a similar fashion; however, the accumulation of protein tyrosine phos-phorylation appeared more pronounced (approxi-mately twofold) The rapid phosphorylation progress and the appearance of discrete bands, as illustrated in Fig 4A, are typical of a processive phosphorylation process [22,23,37], governed by an initial interaction between the Fyn SH3-domain and PAG This binding mode is indispensable for rapid and efficient PAG phosphorylation, as full-length PAG-RLP* with a

PRD1 PRD1-RLP*

PRD1-super PRD2

Wavelength (nm)

2 ·dmol

–1 ) (x10

5 )

50

0

–50

–100

–150

–200

–250

–300

Fig 3 Far-UV CD spectroscopy of the PAG-derived peptide

ligands Spectra were recorded in 10 m M Hepes (pH 7.4), 150 m M

NaCl at a sample concentration of 90 l M.

Time (min)

wt PAG PAG RLP*

super PAG

0 0 50 100 150 200 250 300

75 kDa

0 20 40 60 80 120

Time

100

100 kDa

10 20 min

s

pPAG

A

B

PA G

Fig 4 Kinetics of PAG phosphorylation by wild-type Fyn In vitro pulse-chase assays were performed using active Fyn kinase and various full-length PAG constructs incubated on ice in the presence

of 80 n M [32P]ATP[cP] for 10 min Addition of 0.5 m M ATP initiated the chase reaction, aliquots were withdrawn at specified time points and reactions stopped by mixing with SDS ⁄ PAGE sample buffer and boiling Time points were resolved on SDS ⁄ PAGE gels and detected using autoradiography (A) Phosphorylation of wild-type PAG using active Fyn kinase illustrating the autoradiography gels used for quantification of phosphorylation levels Coomassie stain shows equal loading of PAG in each sample (B) Tyrosine phosphorylation effects of the PAG–Fyn SH3 interaction modula-tion Phosphorylation levels were quantified as pixel intensity of the highest molecular species relative to phosphorylated wild-type (wt) PAG at 5 min Error bars show SEM over three repeat experiments.

greatly reduced Fyn SH3-domain affinity showed a

significantly delayed tyrosine phosphorylation

Phos-phorylation of the PAG-RLP* protein most likely

fol-lows a distributive course caused by random collisions

between the substrate and kinase Efficient

phosphory-lation of PAG-RLP* consequently occurs only after

phosphorylation of Tyr168 and Tyr181, which

accord-ing to our interaction model [22] allows Fyn to dock

onto its substrate via SH2-domain binding, thus

switching to tyrosine phosphorylation in a processive

manner Again, the interaction was deemed specific for

the first proline-rich PAG sequence, as all substrate

molecules included in the kinase assays contained an unaltered second proline-rich region (amino acids 252–

265 of PAG, human) which did not appear to affect tyrosine phosphorylation by Fyn in vitro

Structural impact of Fyn SH3-domain ligand binding

Far-UV CD was employed to characterize local struc-tural changes in the Fyn SH3-domain following binding

of PAG-derived peptide ligands CD spectra of the Fyn SH3-domain alone (Fig 5A) showed a minimum at

A

B

C

Fig 5 Far-UV CD spectroscopy of the Fyn

SH3-domain interaction with PAG derived

ligands (A) Far-UV CD spectra of the Fyn

SH3-domain, the PRD1 peptide and the SH3

domain-peptide complex measured in

10 m M Hepes (pH 7.4), 150 m M NaCl (B)

Spectra of SH3-domain ligand mixtures at a

ligand concentration range of 10–90 l M.

Spectra of free peptide were subtracted

from those of the mixture (C) Difference

spectra of the SH3-ligand mixtures [data

presented in B with spectra of the free

SH3-domain subtracted (a )b)].

207 nm and two maxima at 221 and 236 nm,

compara-ble with previous reports on SFK SH3-domains [38,39]

Addition of peptide containing the wild-type binding

site (PRD1) resulted in a positive shift of the maxima

and an increase in positive ellipticity over the 220–

240 nm region, whereas the minimum was slightly

blue-shifted with an increase in negative ellipticity The latter

observation may indicate an increase in PPII

conforma-tion caused by stabilizaconforma-tion of the proline-containing

peptide upon binding to the SH3-domain [34]

Ligand binding over a concentration range of 10–

90 lm showed that the wild-type sequence and the

high-affinity SH3-binder, PRD1-super, induced a local

concentration-dependent structural change over the

same region (220–240 nm, Fig 5B) Spectra for both

peptides were found to approach a saturation

maxi-mum Increasing concentrations of the low-affinity

SH3-binder, PRD1-RLP*, did not affect secondary

structure, as estimated by CD measurements; neither

did the control peptide PRD2 containing the second

proline-rich region The largest difference, Dh, in the

215–250 nm region was plotted for the highest peptide

concentrations (90 lm, Fig 5C) This clearly revealed

that the relative local structural change was signifi-cantly greater for the high-affinity binder, showing a broad positive band with an ellipticity maximum cen-tred at 225 nm Positive ellipticity in this region has been attributed to interactions with clustered aromatic amino acids [39,40], which may involve tyrosine resi-dues located in the interaction loops of the SH3-domain Other parts of the differential spectrum were not significantly influenced by ligand binding to the SH3-domain for either peptide (data not shown)

Determination of affinity and kinetic constants Binding affinities and dissociation constants for ligands derived from the first proline-rich domain of PAG were analysed using SPR To this end, we immobilized Fyn SH3-fusion proteins to surfaces of biosensor chips, as described above, and assayed for binding by passing the peptides PRD1, RLP* and PRD1-super over immobilized proteins Thus, a single chip was used for all peptides, minimizing effects of varia-tions in ligand concentration Binding profiles (Fig 6A) obtained this way corresponded well with

Fig 6 SPR measurements of the interaction between immobilized Fyn SH3-domains and peptides PRD1, PRD1-RLP* and PRD1-super (A) All three peptides were injected over the sensor chip at 100 l M to reveal differences in binding (B–D) Peptides (B, PRD1; C, PRD1-RLP*; D, PRD1-super) were injected over the sensor chip for kinetic analysis of the SH3-domain interaction Ligand binding curves for peptide concen-trations of 3.1, 12.5, 50 and 200 l M are shown Binding was measured in resonance units (RU), where 1 RU corresponds to the binding of

1 pgÆmm)2.

results from pull-down assays using biotinylated

tides Approximately twice as much PRD1-super

pep-tide as the PRD1 peppep-tide was found to bind to

immobilized SH3-domains over multiple experiments,

while virtually no binding to the immobilised proteins

was observed for the PRD1-RLP* peptide

To determine association and dissociation rate

con-stants for the peptides, concentration series were

injected over the sensor chips (Fig 6B–D,

representa-tive curves from a single run are shown) Kinetic data

from the binding curves were evaluated using a

two-state conformational change model, which was found

to provide a better fit than the 1 : 1 Langmuir model

based on analyses of residuals and v2 values Our

choice of model is backed by findings using CD

spec-troscopy (Fig 5), which indicated that ligand binding

induced a structural change in the Fyn SH3-domain

A summary of kinetic constants is given in Table 1

Dissociation constants showed that the PRD1-super

peptide bound to the Fyn SH3-domain with

approxi-mately fourfold higher affinity than the wild-type

inter-action sequence KDestimated for the wild-type PRD1

peptide was in the low micromolar range ( 4–7 lm)

in repeat experiments, of the same order of magnitude

as previous studies on SFK SH3-domains using similar

technology [28,39,41–43]

Molecular modelling of the Fyn SH3-domain and

PAG-derived peptides

The interaction pockets of the SH3-domain are made

by loops linking the individual b strands together (the

RT loop, the n-Src loop, and the 310helix as indicated

in Fig 1C), flanked by strands b4 and b5 The variable

loops n-Src and RT are principal determinants for

ligand recognition, orientation, and specificity of this

domain, with residues Tyr91 and Tyr137 forming

interaction pocket 1; Tyr93, Tyr137 and Trp119

form-ing pocket 2, and the valley between n-Src and RT

loops included by Trp119 and Tyr132 making up the third interaction pocket [31,44]

Molecular modelling of the ligand⁄ SH3-domain interactions were performed using the modeller soft-ware as described above Three available structures (1fyn [25], 1azg [26] and 1a0n [26]) showing the Fyn SH3 domain binding to peptides were superposed to establish the positioning of the SH3 binding PxxP motif and the alignment of our peptides to a molecular model (Fig 7) To model our peptides binding to the SH3 domain, we selected the NMR structure 1azg of the Fyn SH3 domain complexed to a proline-rich pep-tide P2L (PPRPLPVAPGSSKT) corresponding to resi-dues 91–104 of the p85 subunit [26] P2L contains the class I SH3 consensus motif RxxPxxP, sharing sequence properties with both PRD1 and PRD1-super The first four residues of each peptide (CHQS) were common to all models, thus, all peptide sequences were numbered starting with the first residue of the model sequence, arginine for PRD1 and PRD1-super and alanine for PRD1-RLP*, respectively

Alignment of the PRD1 peptide containing the wild-type PAG interaction sequence binding to Fyn SH3 (Fig 7B), predicted both hydrophobic and electrostatic interactions involved in formation of the complex The model showed peptide residues Pro4 and Pro7 fitting well into binding grooves 1 and 2 created by the n-Src loop and the 310helix, stabilized by stacking interac-tions of the pyrrolidine ring of Pro7 with the aromatic side chain of Tyr137 Modelling of the PRD1-super peptide, which included two amino acid substitutions (Glu2Pro and Pro8Arg), again showed main interac-tions via the core residues for binding (Arg1, Pro4 and Pro7) The proline substitution (Pro2) positioned directly after the arginine residue occupied pocket 3 and was stacked against the aromatic residue Trp119 Previous reports have suggested that pocket 3, binding arginine via acidic side chains, could favourably accommodate binding of a proline residue in this way [45], and it is likely that the proline substitution pro-vides a major contribution to the increased affinity for the PRD1-super peptide Advantages of the Arg8 introduction were less obvious as alignment of flanking residues correlated with greater flexibility in position-ing; however, further stabilization of the complex could be provided by H-bond interactions via water molecules and other amino acids not evaluated in this model

Impact of the various amino acid substitutions was demonstrated by superposing the variant peptides PRD1-RLP* (green) and PRD1-super (blue) onto the wild-type PRD1 peptide (beige) (Fig 7C) The mod-elled structures showed that alanine substitution of

Table 1 Kinetic constants for PAG-derived ligands and Fyn

SH3-domain interactionsa.

k a1

( M )1Æs)1)

k d1

(s)1) ka2(s)1)

k d2

(s)1) KD( M) PRD1 7.7 · 10 4 0.41 2.5 · 10)4 0.01 5.2 · 10)6

PRD1-super 6.8 · 10 4 0.1 6.9 · 10)4 0.003 1.2 · 10)6

a

Apparent association (k a ) and dissociation (k d ) rate constants and

affinity constants (KD) were calculated from three independent

experiments Numbers in the table are shown for one

representa-tive experiment No kinetic data could be obtained for the Fyn

SH3 ⁄ PRD1-RLP* peptide interaction SD < 1% for all constants,

apart from PRD1 ka2and kd2where SD < 5%.

Arg1 appeared to reduce interactions with acidic

resi-dues of pocket 3 (e.g resiresi-dues surrounding Trp119)

The substitution removes the arginine residue of the

RxxPxxP motif, thus abrogating the formation of the

salt bridge with Asp99, which defines the orientation

of the peptide ligand [33] Alanine substitution of

Pro4 eliminated its stacking interactions with the

cen-tral tryptophan residue Trp119, believed to

exten-sively reduce the ability of the PRD1-RLP* ligand to

form a stable complex with the Fyn SH3-domain

The overlay structures emphasised the significance of

the PRD1-super Pro2 substitution, revealing a good

fit adjacent to Arg1 where interactions with aromatic

and acidic residues such as Trp119, Asp99 and

Asp118 of the SH3-domain were likely It is highly

probable that the predicted interactions outlined

above result in the local structural changes in the

Fyn SH3-domain as observed by CD measurements,

and that these changes affect binding affinity as

eval-uated by SPR

Conclusions

We have described in detail the interactions of a nat-ural Fyn SH3-domain ligand extracted from the adap-tor protein PAG and low- and high-affinity variants of this sequence developed using a 2D array approach Interaction kinetics and local structural impact of these ligands binding to the Fyn SH3-domain were analysed using SPR and CD, respectively, and the results of these investigations related to functional effects of ligand binding, assessed experimentally by kinase assays using full-length PAG proteins containing the modified binding motifs as substrates In summary, our data demonstrate that substrate phosphorylation may be modified through minor changes in the SH3-domain interacting motif

Optimal binding sequence motifs have previously been found using random or biased library approaches, performed with several cycles of selection for optimalization [27–29,46] In this study, we used a

Fig 7 Molecular models showing the Fyn SH3-domain complexed to PAG-derived peptides (A) Superposition of the three available PDB structures for the Fyn SH3 domain complexed to ligand peptides (1azg (yellow) [26], 1fyn (green) [25] and 1a0n (red) [26]) (B) Models of the Fyn SH3-domain in complex with PRD1 and PRD1-super peptides based on the NMR structure 1azg [26] The peptide is displayed

as sticks; the SH3-domain is shown as rib-bons, while selected amino acids predicted

to be essential for interaction with the ligand are shown as spheres (C) Superposi-tion of the RLP* (green) and PRD1-super (blue) peptides on the PRD1 (beige) peptide complexed to the Fyn SH3-domain.

natural ligand as a point of reference for the

develop-ment of high- and low-affinity ligands using a 2D

pep-tide array approach This is a simple and rapid

method for analysis of individual amino acid

substitu-tions, and the high-affinity SH3-binder, PRD1-super,

developed using this strategy was found to have a

dis-sociation constant in the low micromolar range,

com-parable to that of the phage-display library peptide

VSL12 (1.2 versus 0.6 lm) [46]

In vivo manipulation of protein interaction domains

in this way could provide valuable insight into the

functional consequences of localized protein–protein

interactions As an example, the high-affinity,

full-length PAG (superPAG) construct has been used in

T cells to reveal that the negative regulatory potential

of PAG is enhanced by this modification, consistent

with a higher degree of PAG phosphorylation and

concomitant Csk recruitment [22] On a similar note, is

the kinase targeting strategy used by viruses such as

HIV and Herpesvirus saimiri which encode accessory

proteins containing proline-rich domains [47–49]

These proteins, Nef and Tip, respectively, interact with

kinase SH3-domains with high affinity to modify the

behaviour of virus-infected cells, and have been shown

to control both pathogenicity and T-cell proliferation

This demonstrates the potential of manipulating

signal-ling pathways in vivo by creating high-affinity ligands

to compete with natural SH3-protein interactions

Experimental procedures

Peptide synthesis

Peptides PRD1 (CHQSRELPRIPPES), PRD1-RLP*

(CHQ-SAEAARIPPES), PRD1-super (CHQSRPLPRIPRES) and

PRD2 (EEEAPPPVPVKLLD) were synthesized with or

without biotin-tags coupled to the N-terminus in-house or

purchased from AnaSpec Inc (San Jose, CA, USA) Purity

was analysed by HPLC and mass spectroscopy All

and peptide concentration was determined using a

Bio-chrom 30 amino acid analyser (BioBio-chrom, Cambridge,

UK)

Expression plasmids of glutathione

S-transferase -tagged proteins

The constructs expressing different variants of glutathione

S-transferase (GST)-PAGDTM have been described

previ-ously [22] A construct expressing the SH3-domain of Fyn

was created by sub-cloning amino acids 83-537 into the

intro-ducing a stop codon at amino acid residue 142

Biotin-peptide pull-down assay

Human peripheral blood T cells were purified from normal donors by negative selection as described [18] and lysed in ice-cold lysis buffer (50 mm Tris pH 7.4, 100 mm NaCl,

phenylmethylsulfonyl fluoride) After 30 min pre-clear at

USA), T-cell lysates were incubated with different

30 min incubation with streptavidin beads at room temper-ature Biotin alone was used as a control for non-specific binding Samples were washed extensively in ice-cold lysis buffer prior to western blot analysis using anti-Fyn IgG (Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA)

Protein expression and purification

Recombinant proteins were expressed and purified as described previously [22] For thrombin cleavage of the GST-tag, washed glutathione–Sepharose beads bound to GST–Fyn SH3-domain were incubated with human throm-bin (Sigma, St Louis, MO, USA) in 50 mm Tris (pH 8.0),

SH3-domain was purified further by gel filtration over a Superdex

75 column (GE Healthcare Europe GmbH, Uppsala, Sweden) using 10 mm Hepes (pH 7.4), 150 mm NaCl, and concentrated using a Vivaspin 2 spin column (Sartorius Bio-lab, Auckland, New Zealand) Recombinant proteins were

In vitro kinase assay

Pulse-chase experiments using full-length, active Fyn kinase (Upstate, Charlottesville, VA, USA) were performed as described elsewhere [37] and protein phosphorylation quan-tified using imagequant (GE Healthcare Europe)

Circular dichroism

CD spectroscopy was performed on a nitrogen-flushed JASCO spectropolarimeter J810 (Jasco, Tokyo, Japan) equipped with a circulating water bath Samples were

Mu¨llheim, Germany) with a pathlength of 0.1 cm CD

1 nm over a wavelength range of 190–260 nm Samples were analysed in 10 mm Hepes, pH 7.4, 150 mm NaCl Total buffer absorbance was analysed and was found to be satisfactory for the range of wavelengths used in the measurements The scans were averaged and a sample-free buffer spectrum was subtracted Data smoothing was

performed using an inverse square algorithm in sigmaplot

8.0 (SPSS, Chicago, IL, USA) as described previously [50]

Surface plasmon resonance

Measurements were performed on a BIAcore T100

GST–Fyn SH3 fusion protein or GST protein alone

(diluted in 10 mm sodium acetate, pH 4.5) were

immobi-lized on CM5 sensor chips (BIAcore) via cross-linking of

1-ethyl-3-[3-dimethylaminopropyl] carbodiimide hydrochloride

acti-vated flow-cell surfaces to a final response of 8500 RU

succini-mide ester groups with 1 m ethanolamine GST–protein was

immobilised as a reference surface to the same molar

extent The chip was treated with 10 mm dithiothreitol in

running buffer (10 mm Hepes, pH 7.4, 150 mm NaCl,

0.05% P20, 3 mm EDTA, 0.005% SDS) to remove GST

dimer products After extensive washing of the surface with

running buffer, peptide binding was assessed by injecting

the indicated concentrations in running buffer over the

analysis was performed using the BIAcore T100 evaluation

software Curve fitting was executed by the software

pro-gram (global fitting algorithm) which also performed

Molecular modelling of the protein–peptide

complexes

Modelling of the protein–peptide complexes was performed

using modeller [51] and visualized using pymol (http://

www.pymol.org) The Protein Data Bank [52] entry of the

Fyn tyrosine kinase SH3-domain bound to a proline-rich

peptide 3BP-2 (PPAYPPPPPVP, PDB ID: 1fyn [25]) and

P2L (PPRPLPVAPGSSKT, PDB ID: 1azg and 1a0n [26])

were used to establish the positioning of the PxxP core

motif of the peptides and 1azg was selected as a template

to create models for binding of the PRD1, PRD1-super,

and PRD1-RLP* peptides In aligning the peptides of this

study to that in the 1azg, it was assumed that the core

SH3-domain binding motif (PxxP) would overlay with the

equivalent position as described in the original study [26]

Acknowledgements

This work was supported by Norwegian Research

Council FUGE Career Fellowship (159306 to TB), by

EU Grant 3D repertoire LSHG-CT-2005-212028 (to

RR), and by grants from the Norwegian Functional

Genomics Programme, the Norwegian Research

Council, the Norwegian Cancer Society, and the

European Union (grant no 037189, thera-cAMP) (to

KT) We thank Gladys M Tjørhom, Jorun Solheim, and Ola Blingsmo for excellent technical assistance,

Dr Per E Kristiansen for use of the circular dichro-ism equipment, and Dr Matthew Betts for helpful discussion

References

1 Pawson T (2004) Specificity in signal transduction: from phosphotyrosine–SH2 domain interactions to complex cellular systems Cell 116, 191–203

2 Hubbard SR & Till JH (2000) Protein tyrosine kinase structure and function Annu Rev Biochem 69, 373–398

3 Palacios EH & Weiss A (2004) Function of the Src-fam-ily kinases, Lck and Fyn, in T-cell development and activation Oncogene 23, 7990–8000

4 Gschwind A, Fischer OM & Ullrich A (2004) The dis-covery of receptor tyrosine kinases: targets for cancer therapy Nat Rev Cancer 4, 361–370

5 Hermiston ML, Xu Z & Weiss A (2003) CD45: a criti-cal regulator of signaling thresholds in immune cells Annu Rev Immunol 21, 107–137

6 Hunter T (1995) Protein kinases and phosphatases: the yin and yang of protein phosphorylation and signaling Cell 80, 225–236

7 Pao LI, Badour K, Siminovitch KA & Neel BG (2007) Nonreceptor protein-tyrosine phosphatases in immune cell signaling Annu Rev Immunol 25, 473–523

8 Boggon TJ & Eck MJ (2004) Structure and regulation

of Src family kinases Oncogene 23, 7918–7927

9 Xu W, Harrison SC & Eck MJ (1997) Three-dimen-sional structure of the tyrosine kinase c-Src Nature 385, 595–602

10 Guappone AC & Flynn DC (1997) The integrity of the SH3 binding motif of AFAP-110 is required to facilitate tyrosine phosphorylation by, and stable complex forma-tion with, Src Mol Cell Biochem 175, 243–252

11 Nakamoto T, Sakai R, Ozawa K, Yazaki Y & Hirai H (1996) Direct binding of C-terminal region of p130Cas

to SH2 and SH3 domains of Src kinase J Biol Chem

271, 8959–8965

12 Richard S, Yu D, Blumer KJ, Hausladen D, Olszowy

MW, Connelly PA & Shaw AS (1995) Association of p62, a multifunctional SH2- and SH3-domain-binding protein, with src family tyrosine kinases, Grb2, and phospholipase C gamma-1 Mol Cell Biol 15, 186–197

13 Straus DB & Weiss A (1992) Genetic evidence for the involvement of the lck tyrosine kinase in signal trans-duction through the T cell antigen receptor Cell 70, 585–593

14 Yasuda K, Nagafuku M, Shima T, Okada M, Yagi T, Yamada T, Minaki Y, Kato A, Tani-Ichi S, Hamaoka

T et al (2002) Cutting edge: Fyn is essential for