DOI 10.1007/s00702-006-0515-5

J Neural Transm (2006)

Biomolecular interaction analysis in functional proteomics

D. Moll1, A. Prinz1, F. Gesellchen1, S. Drewianka2, B. Zimmermann2,and F. W. Herberg1

1 Department of Biochemistry, University of Kassel, and2 Biaffin GmbH & Co KG, Kassel, Germany

Received September 25, 2005; accepted April 5, 2006Published online July 13, 2006; # Springer-Verlag 2006

Summary. To understand the function ofhighly complex eukaryotic tissues like thehuman brain, in depth knowledge about cel-lular protein networks is required. Biomole-cular interaction analysis (BIA), as a part offunctional proteomics, aims to quantify inter-action patterns within a protein network indetail. We used the cAMP dependent proteinkinase (PKA) as a model system for the bind-ing analysis between small natural ligands,cAMPand cAMPanalogues, with their physio-logical interaction partner, the regulatory sub-unit of PKA. BIA comprises a variety ofmethods based on physics, biochemistry andmolecular biology. Here we compared side byside real time SPR (surface plasmon resonance,Biacore), a bead based assay (AlphaScreen),a fluorescence based method (Fluorescencepolarisation) and ITC (isothermal titrationcalorimetry). These in vitro methods werecomplemented by an in cell reporter assay,BRET2 (bioluminescence resonance energytransfer), allowing to test the effects of cAMPanalogues in living cells.

Keywords: cAMP dependent protein kinase,surface plasmon resonance, fluorescencepolarization, isothermal titration calorimetry,bioluminescence resonance energy transfer,AlphaScreen.

Abbreviations

AlphaScreen amplified luminescence proxi-mity homogeneous assay; BIA biomolecularinteraction analysis; BRET bioluminescenceresonance energy transfer; cAMP adenosine-30,50-cyclicmonophosphate; C catalytic sub-unit of PKA; FP fluorescence polarization;ITC isothermal titration calorimetry; PKAcAMP dependent protein kinase; PTM posttranslational modifications; R regulatory sub-unit of PKA; RU response units; SPR surfaceplasmon resonance.

Introduction

The function of biological systems is me-diated by proteins and their interactions. Cel-lular activities are not only dependent onprotein expression patterns, but are also con-trolled by post translational modifications(PTMs), by compartmentalization and proteindegradation (Graves and Haystead, 2002). Inthe classical proteomics approach methodslike 2D gel electrophoresis and mass spec-trometry have been established to describeexpression patterns differentially e.g. compar-ing healthy and diseased tissue. One majorgoal of functional proteomics is to determineprotein–protein, protein-DNA, and protein-

ligand interaction with high accuracy and ingreat detail within a cellular network. Intra-cellular interaction pathways are influencedto a large extent by PTMs (Shaywitz et al.,2002; Yaqub et al., 2003). The most promi-nent PTM in an eukaryotic cell is proteinphosphorylation, a key event in the regula-tion of the cell mediated by the action of pro-tein kinases (Manning et al., 2002). About40% of all proteins may undergo this crucialPTM during some state of cellular growth anddifferentiation. The importance of proteinphosphorylation in signalling pathways is im-pressively reflected in the devastating effectsof protein kinase dysfunction linked to severalhuman diseases (Blume-Jensen and Hunter,2001; Sachsenmaier, 2001) most prominent-ly in human malignancies (Fabbro et al.,2002; Tasken and Aandahl, 2004). About518 Serine=Threonine and Tyrosine specificprotein kinases are encoded in the humangenome (Manning et al., 2002). Although pro-tein kinases differ substantially in their sub-strate specificity, activity, biological half life,cellular localisation and function, they sharea common overall protein fold. Cyclic AMPdependent protein kinase (PKA) has beenused as a model system for kinase functionduring the last two decades and has been char-acterized biochemically and structurally indetail (Taylor, 1989; Knighton et al., 1991;Taylor et al., 2004). The enzyme consists ofa regulatory (R) dimer and two monomericcatalytic (C) subunits forming an inactive ho-loenzyme complex (R2C2). The holoenzymecomplex is activated by increasing concen-trations of the second messenger cAMP, alow molecular weight ligand. Upon bindingof cAMP to two distinct binding sites on eachR subunit, the C subunits are released (Fig. 1)and can phosphorylate substrates in the cyto-plasm, but they can also migrate into thenucleus, thus affecting gene regulation viathe cAMP response element binding protein(CREB (Hagiwara et al., 1993)). Not surpris-ingly, deregulation of PKA activity has beenimplicated in various human diseases, among

them breast cancer, Carney complex andHIV infection (Tasken and Aandahl, 2004).PKA has received increasing attention as afactor contributing to neurological disorders.It has already been known that PKA activityis required for memory formation in themammalian hippocampus (Kandel, 2001).Recently, it was shown that PKA is alsoinvolved in processes concerning the workingmemory as well as reward-motivated learn-ing, linking PKA action potentially to mem-ory deficits and drug addiction (for review seeArnsten et al., 2005).

Due to its modular structure the PKA sys-tem is ideally suited as amodel for the analysesof protein–protein interaction and protein-small ligand interaction. Here we analyzethe binding of the regulatory subunit of PKAto both the natural small ligand cAMP and tothe catalytic subunit.

Several methods based on different phy-sico-chemical parameters are available to de-termine protein–protein interactions. Thesetechniques summarized under the term bio-molecular interaction analysis (BIA), allowan in depth look at molecular interactions.In the following we will describe interac-tions within the PKA model system deter-mined by surface plasmon resonance (SPR),AlphaScreen, fluorescence polarization (FP),isothermal titration calorimetry (ITC) andbioluminescence resonance energy transfer(BRET2). These methods comprise solid phase

Fig. 1. Model of PKA holoenzyme activation. Theinactive holoenzyme of cAMP dependent protein ki-nase (PKA) consists of a regulatory subunit dimer (Rsubunit, dark grey) and two catalytic subunits (C sub-unit, light grey). For simplification, only one C subunitinteracting with one R subunit monomer is shown.Binding of two molecules of cAMP (filled circle) toan R subunit monomer leads to the dissociation of theholoenzyme complex, thereby releasing the active C

subunit

D. Moll et al.

assays as well as homogenous assays, per-formed in vitro and in living cells.

Materials and methods

Reagents

The synthetic peptide substrate Kemptide (LRRASLG)was purchased from Biosyntan GmbH (Berlin,Germany). ATP and NADH were obtained from Bio-mol GmbH (Hamburg, Germany).

P11 cation exchanger cellulose and DE52 (Diethy-laminoethyl, DEAE) anion exchanger cellulose wereobtained from Whatman (Maidstone, UK).

cAMP (adenosine-30,50-cyclicmonophosphate),8AHA-cAMP (8-(6-aminohexyl)aminoadenosine-cAMP), 8CPT-cAMP (8-(4-chlorophenylthio)-cAMP),2Cl-cAMP (2-chloroadenosine-cAMP), 6AH-cAMP(N6-(6-aminohexyl)-cAMP), 8Fluo-cAMP� (8-[[2-[(fluoresceinylthioureido)amino]ethyl]thio]-cAMP),6Ph-cAMP (N6-phenyl-cAMP), 6MAH-cAMP (N6-(6-[N0-methylanthraniloyl]aminohexyl)-cAMP), 8Br-cAMP (8-bromo-cAMP), 8Cl-cAMP(8-chloro-cAMP),6MB-cAMP (N6-monobutyryl-cAMP), 8NBD-cAMP�

(8-[[2-[(7-nitro-4-benzofurazanyl)amino]ethyl]thio]-cAMP), 8ADOA-cAMP (8-(8-amino-3,6-dioxaoctyla-mino)-cAMP), Sp8AHA-cAMP (8-(6-aminohexyl)aminoadenosine-30,50-cyclicmonophosphorothioate, Sp-isomer), 8PIP-cAMP (8-piperidino-cAMP), 8AEA-cAMP (8-(2-aminoethyl)amino-cAMP), 2AHA-cAMP(2-(6-aminohexyl)amino-cAMP), 20Mant-cAMP� (20-O-(N-methylanthraniloyl)-cAMP), 8Br-cAMP-AM(8-Bromoadenosine-30,50-cyclicmonophosphate, aceto-xymethylester), Sp8Br-cAMPS (8-bromoadenosine-30,50-cyclicmonophosphorothioate), cGMP (guanosine-30,50-cyclicmonophosphate) and cUMP (uridine-30,50-cyclicmonophosphate) were obtained from Biolog LifeScience Institute (Bremen, Germany). Fine chemicals(research grade) were purchased from Roth (Karlsruhe,Germany) or from Sigma-Aldrich (Deisenhofen,Germany).

CM5 sensor chips (research grade), NHS (N-hydro-xysuccinimide), EDC (N-ethyl-N0-(dimethylaminopro-pyl)-carbodiimide), ethanolamine-HCl, and surfactantP20were obtained fromBiacore AB (Uppsala, Sweden).

Preparation of recombinant proteins

The cDNAs for the expression of recombinant pro-teins were kind gifts from Prof. S. S. Taylor, Universityof California, San Diego, USA (murine Ca and bovineRIa�1-91 (RI monomer)) and Prof. K. Tasken, Univer-

sity of Oslo, Oslo, Norway (GST-hRIa and hRII

a ).RI monomer was overexpressed in E. coli BL21

(DE3) and purified according to Herberg et al. using

ion exchange chromatography (Herberg et al., 1994).GST-hRI

a was purified using Glutathion agarose fromSigma (Deisenhofen, Germany) following standardprotocols (Sambrook et al., 2001).

To obtain cAMP free R subunits, the purified pro-tein was incubated with 10mM cGMP over night at4�C. Subsequently, excess cGMP was removed usinga PD 10 desalting column (Amersham Biosciences,Freiburg, Germany) followed by extensive dialysisagainst 150mM NaCl, 20mM MOPS, pH 7 (bufferA). To obtain completely nucleotide-free R subunitfor ITC measurements, the purified RI monomer wastreated with 8M urea to remove cAMP and then dia-lyzed as described above (Buechler et al., 1993).

Recombinant PKA C subunit was expressed andpurified as described before (Slice and Taylor, 1989;Herberg et al., 1993).

The purity of the R subunits was confirmed bySDS-polyacrylamide gel electrophoresis and the biolo-gical activity of the proteins was verified using thephosphotransferase assay with the peptide Kemptide asa substrate according to Cook et al. (1982).

AlphaScreen

Biotin labelling of C subunit was performed with a 10:1molar excess of EZ-Link NHS Biotin (Perbio Sciences,Bonn, Germany) according to the manufacturer’s in-structions. The reaction was performed with intactholoenzyme, in order to protect the C subunit=R sub-unit interface from chemical modification during thebiotinylation procedure. After biotinylation, cAMP wasadded to the reaction mixture to dissociate the holo-enzyme complex and the free biotinylated C subunitwas subsequently purified using PKI(5–24) affinitychromatography (Olsen and Uhler, 1989).

For the actual AlphaScreen measurements, biotiny-lated C subunit and GST-tagged R subunit (0.2 nM each)were mixed together in the presence of serial dilutionsof cAMP or Sp8Br-cAMPS in a 384 well microtiterplate (Optiplate, white, PerkinElmer). In a followingstep anti-GST acceptor beads and streptavidin donorbeads (PerkinElmer, Rodgau, Germany) were added toa final concentration of 20mg=ml. The final reactionvolume in each well was 25ml in 25mM Hepes, pH7.4, 100mM NaCl, 10mM MgCl2, 1mM ATP, 0.1%BSA. The resulting AlphaScreen signal (counts persecond) was determined with a FusionTM a-FP micro-titerplate reader (Packard Bioscience, now PerkinElmer)after one hour incubation at room temperature.

Fluorescence polarization (FP)

Two different assay formats were used. In the directassay format both the RI monomer concentration aswell as the 8Fluo-cAMP concentration were varied

Biomolecular interaction analysis

(10 pM to 1mM for RI monomer and 20nM to 500 pMfor 8Fluo-cAMP). The assay was performed at 20�Cin buffer A containing 0.005% (v=v) CHAPS as asurfactant in a 384 well microtiterplate (Optiplate,black) using the FusionTM a-FP microtiterplate reader.The fluorescence polarization signal was detected atEx 485 nm=Em FP Filter 535 nm with a PMT Voltageof 1100.

For the fluorescence polarization displacementassay increasing concentrations of cAMP analogues(typically ranging from 1 fM to 1mM) were mixed with1 nM 8Fluo-cAMP before adding the RI monomer.Here the concentration of the RI monomer was adaptedto 80% of the maximum value derived from the directassay using 1 nM 8Fluo-cAMP. Fluorescence polariza-tion was measured after 5 minutes. Data were analyzedwith GraphPad Prism 4.0 (GraphPad Software, SanDiego, CA) by plotting the resulting polarization signalagainst the logarithm of the analogue concentration.

Surface plasmon resonance (SPR)

All SPR interaction analyses were performed at 20�Cin buffer A plus 0.005% (v=v) surfactant P20 using aBiacore 3000 instrument (Biacore AB, Sweden). Forcovalent coupling of 8AHA-cAMP, carboxymethylatedsensor chip surfaces (CM5, research grade) were acti-vated with NHS=EDC for 10 minutes and 8AHA-cAMP (3mM in 100mM HEPES, pH 8.0) was injectedfor 7 minutes with a flow rate of 5ml=min. Deactivationof the surface was performed using 1M ethanolamine-HCl (pH 8.5) for 7 minutes. A reference cell wasactivated accordingly and deactivated subsequently.Competition analyses were performed by injection of5 nM RI monomer preincubated with varying concen-trations of each analogue to be analyzed. The bindingsignal was monitored for 3 minutes and data points werecollected at the end of the association phase. The sensorsurfaces were regenerated after each binding cycle byseveral injections of 3M guanidinium HCl. After sub-tracting the reference cell, the resulting binding signalswere plotted against the logarithm of each cAMP ana-logue concentration and an EC50 value was calculatedfrom the dose response curve using GraphPad Prism3.01 (GraphPad Software, Inc., San Diego=USA).

Isothermal titration calorimetry (ITC)

The interaction between RI monomer and cGMP wasanalyzed in buffer A containing 1mM b-mercaptoetha-nol using a VP-ITC microcalorimeter (MicroCal LLC.,Northampton, MA, USA). 5mM RI monomer wereallowed to equilibrate in a 1.385ml cell at 20.0�C. Toensure that the titrant concentration was at its loadingvalue, two injections of 51.3mM cGMP (1ml each) wereperformed before the actual titration experiment. 5ml

injections were conducted until no further binding wasobserved. Each time the reaction returned to baselinelevel (approximately after 3 minutes), a new injectionwas carried out. In order to minimize artifacts, cGMPwas dissolved in exactly the same buffer that was usedto dialyze the purified protein. For blank subtractioncGMP was injected in identical steps into buffer only.Data evaluationwas performedwith the softwareMicro-Cal Origin for ITC (MicroCal LLC., Northampton, MA,USA; see also Wiseman et al., 1989), including correc-tions for volume change during the titration.

BRET2 assay

The human Ca coding sequence was amplified usingsense and antisense primers harboring unique HindIIIand BamHI sites. The fragment was subcloned in-frameinto the HindIII=BamHI sites of pGFP2-C vector(PerkinElmer). The human RII

a coding sequence wasamplified without the stop codon allowing for cloningwith BamHI and KpnI. The fragments were subclonedusing the pTrcHis2-Topo+ TA cloning kit (Invitrogen,Karlsruhe, Germany), excised and cloned in-frame intoBamHI=KpnI sites of pRluc-N vector (PerkinElmer).

For BRET2 assays, COS-7 cells were seeded ina white 96 well microtiterplate (Optiplate, White,PerkinElmer) at a density of 2�104 cells per well.Transfections were carried out in the microtiterplate24 hours later using 4ml Polyfect reagent (Qiagen N.V.,Venlo, Netherlands) and a 0.5mg plasmid DNA perwell and transfection. 48 hours post transfection, cellswere washed with glucose-supplemented Dulbecco’sPBS (D-PBS, Invitrogen), and the substrate Deep-BlueC(tm) (PerkinElmer) was added to a final concen-tration of 5mM in a total volume of 50ml D-PBS. Lightemission was detected using a FusionTM a-FP mi-crotiterplate reader (PerkinElmer; read time 1 second,gain 50). The light output was taken consecutively foreach well using filters at 410 nm wavelength (�80nmbandpass) for the donor and 515 nm (�30nm band-pass) for the acceptor fluorophore. Background values,routinely obtained with untransfected cells, were sub-tracted in each measurement. Control transfections withempty pRluc and pGFP2 vectors were carried outwith each experiment. BRET2 signals were calculatedas (emission 515 nm-background515nm)=(emission 410 nm-background410nm). GraphPad Prism 4.0 was used forstatistical analysis.

Results

Biomolecular interaction analysis (BIA) isused to describe binding events between smallmolecules, proteins (e.g. receptors, enzymesand antibodies), peptides, nucleotides or car-

D. Moll et al.

bohydrates. Here we focus on one hand on in-teraction between the PKA regulatory subunitand its low molecular weight ligand cAMPand on the other hand on the influence ofcAMP on holoenzyme formation using fourin vitro methods and one in cell assay. Equi-librium binding data as well as associationand dissociation rate constants were obtained.

AlphaScreen allows compound screeningin a homogenous assay format

AlphaScreen is a bead-based proximity assay,where a luminescence signal is the readout fora biomolecular interaction. Both interactionpartners have to be chemically coupled tolatex beads (diameter 250 nm). Upon illumi-nation with laser light (lex¼ 680 nm) singletoxygen ðO2ð1�gÞÞ is produced on the donor

beads (D), see Fig. 2 Inset. If a second inter-action partner, coupled to acceptor (A) beads,binds, both beads are brought into close prox-imity. The chemical energy of the singlet oxy-gen from the donor beads can diffuse to theacceptor beads, resulting in a luminescencesignal (lem¼ 520–620 nm). If the interactionpartners do not bind to each other, the singletoxygen decays (t1=2¼ 4ms) and no lumines-cence signal is observed. The assay is usuallyperformed in a volume of 5–25ml in 384 wellplates and is well suited for automation andthus for high throughput screening.

Immobilization of the interaction partnersto donor and acceptor beads, respectively, canbe achieved by different strategies. Direct cou-pling of ligands containing primary aminescan be done via reductive amination. However,in the case of PKA R and C subunits this

Fig. 2. Detection of PKA holoenzyme dissociation using AlphaScreen. Biotinylated C subunit and GST-tagged Rsubunit (0.2 nM each) were incubated in a 384 well plate with increasing concentrations of cAMP and Sp8Br-cAMPS, respectively, before adding anti-GST acceptor beads and streptavidin coated donor beads (20 mg=ml finalconcentration). Readings were taken after one hour incubation. Data were normalized and fitted according to asigmoidal dose-response model and EC50 values were calculated. All data points represent the mean� SD oftriplicate measurements. Inset: Depicted in the cartoon is the principle of an AlphaScreen assay. Donor (D) beadswere coupled with C subunits (light grey symbol), acceptor (A) beads with R subunits (dark grey symbol). Dbeads, containing a photosensitizer, release singlet oxygen ðO2ð1�gÞÞ upon illumination with laser light (lex680 nm). Due to the interaction of C subunits coupled to the donor beads and R subunits coupled to the acceptorbeads the chemical energy of the singlet oxygen is converted into a luminescence signal (lem 520–620 nm). If,

upon addition of cAMP, the C and R subunits dissociate, the AlphaScreen signal diminishes

Biomolecular interaction analysis

proved to be difficult, potentially due to stericreasons or a loss of activity during the cou-pling procedure (data not shown). Therefore,an indirect immobilization approach wasused, in which biotin labelled C subunit wascaptured to streptavidin coated donor beads,while a GST-R subunit fusion protein wasimmobilized to donor beads via an anti-GST antibody. Special care had to be takenduring the biotinylation of the C subunit, inorder to keep the crucial interaction site withthe R subunit interaction free from chemi-cal modifications (see material and methods).In the absence of the activator cAMP thisassay setup lead to the generation of a robustluminescence signal reflecting holoenzymeformation by the interaction of C and R sub-units. The signal was reduced to backgroundlevels by increasing cAMP concentrations upto 10 mM, indicating the complete dissocia-tion of the PKA holoenzyme (see Fig. 2).The EC50 values derived from the dose-re-sponse experiments were slightly higher thanthose obtained with other methods (see SPRand FP). Screening of cAMP analogues dem-onstrated that their relative potencies werecorrectly reflected and, consistent withinthe assay (Fig. 2), the EC50 value of Sp8Br-cAMPS was about an order of magnitudehigher than the EC50 value of cAMP.

FP provides a fast and easy setupfor high throughput screening

Fluorescence polarization (FP) is a widelyused optical method in a homogenous assayformat which allows high throughput screens.

It is generally considered to be inexpensiveand rapid with a sensitivity close to a classi-cal radioligand binding assay without the needto separate the bound and unbound ligand(Burke et al., 2003; Jameson and Mocz,2005). The theoretical principles underlyingpolarization measurements have been de-scribed previously (Lakowicz, 1999; Valeurand Brochon, 2001; Jameson et al., 2003;Jameson and Mocz, 2005). Briefly, the basicconcept of fluorescence polarization is that afluorophore is preferentially excited when itis oriented along the electric vector of thepolarized incident light. If the fluorophoredoes not move, the emitted light is still inthe same plane of orientation and thereforeremains polarized. Typically, in a waterysolution, a fluorophore can rotate freely andtherefore the emitted light is no longer polar-ized. In the experimental setup used here afluorescence polarization signal can only bedetected from fluorophores slowed downdrastically in their rotational speed. Thisoccurs for example upon binding of a smallfluorescently labelled ligand to a significantlylarger molecule. The degree of polarization isdetermined by measuring the fluorescenceintensities of parallel and perpendicularemitted light with respect to the plane of lin-early polarized excitation light.

For the optimization of the assay condi-tions, the RI monomer concentration wastested over a range of 1 pM to 100 nM withsix fixed concentrations of 8Fluo-cAMP.

At higher concentrations (�5 nM) 8Fluo-cAMP is titrated directly to the two RI mono-mer binding sites. In this case the EC50 values

1

Fig. 3. Fluorescence polarization. A Direct binding of 8Fluo-cAMP to RI monomer. RI monomer was seriallydiluted in buffer A containing 0.005% CHAPS in a 384 well microtiterplate. 8Fluo-cAMP was added to eachwell using the concentration indicated on the plot. Each data point represents the mean� SEM from triplicatemeasurements. Inset: Schematic illustration of a FP assay where fluorescently labelled cAMP (dots with stars)and unlabeled cAMP (black dots) compete for binding to the R subunit (dark grey symbol). Only labelledcAMP bound to the R subunit generates a FP signal (black arrow). B Displacement experiment. RI monomer(5 nM) was added to a serial dilution of the cAMP analogues indicated on the plot and apparent EC50 werecalculated. For experimental details see materials and methods. Each data point represents the mean� SEM

from six individual wells

D. Moll et al.

Biomolecular interaction analysis

reflect half of the 8Fluo-cAMP concentrationconsistent with the biological model. At con-centrations below 5 nM, equilibrium bindingmeasurements can be performed, reflected bythe fact, that the apparent EC50 value did notchange any further when the ligand concentra-tion was reduced (Fig. 3A). The transitionbetween the titration process and equilibriumbinding conditions can also be seen in a changeof the Hill slope (Hill slope around 1.8 downto concentrations of 5 nM 8Fluo-cAMP and1.2 for concentrations below 5nM).

For competitive binding experiments(Fig. 3B), protein and fluorescent ligand con-centrations were held constant. From the directassay (Fig. 3A) the best signal to noise ratiowas obtained with 1 nM 8Fluo-cAMP. 5 nMRI monomer, corresponding to 80% of maxi-mum signal in the direct assay provide a suffi-cient dynamic range for the competitive assay.Three PKA specific agonists (6AH-cAMP,8AHA-cAMP and 2AHA-cAMP) and un-modified cAMP were tested. The potency ofa given analogue is directly related to the ob-served EC50 value as shown in Fig. 3B. cAMPand 8AHA-cAMP display nearly identicalEC50 values (4 nM and 5.4 nM, respectively),whereas 2AHA-cAMP binds with lower affi-nity (27.8 nM). 6AH-cAMP has the highestaffinity with an EC50 value of 1.4 nM.

SPR generates high quality bindingdata in a wide dynamic range

Although surface plasmon resonance (SPR)based biosensors are commonly used to

directly determine the respective associationand dissociation rate constants of a giveninteraction, this method was expanded hereto a solution competition assay format toanalyze the relative binding affinities of sev-eral cyclic nucleotide derivatives. CyclicAMP analogues containing an aminohexyl-linker either on the 2-, 6- or 8-position ofthe adenine ring were covalently immobi-lised to carboxymethylated dextran surfacesof a CM5 sensor chip at high ligand density.A fixed concentration of R subunit wasinjected over the cAMP surfaces in the pre-sence or absence of the respective cAMPderivative as a competitor in solution. Thechange in SPR signal was monitored inresponse to various concentrations of a com-petitor (Fig. 4B). Since the cAMP ligandswere immobilised in very high densities,the binding of the R subunit becomes mainlymass-transfer limited, reflected in a linearincrease of the binding signal with time.

This solution competition approachallows an automated determination of rela-tive affinities for a whole set of cyclic nucleo-tides binding to different R subunit isoforms.Here, a set of selected cyclic nucleotideswas analyzed for binding to the RI monomerand apparent EC50 values were determined(Fig. 4B inset). The calculated EC50 valuesare displayed in a radar plot according totheir relative affinities (Fig. 4C) ranging fromsubnanomolar values (8CPT-cAMP: 800 pM)via the nanomolar range (cAMP: 1.7 nM) tomicromolar values (cXMP; 50 mM, not de-picted on the plot).

1

Fig. 4. Screening of cAMP analogues in a SPR solution competition assay. A Schematic illustration of a solutioncompetition assay performed on a SPR sensor (Biacore setup). cAMP analogues (black dots) compete withimmobilized cAMP (black dots linked to surface) for binding to R subunit monomers (dark grey symbol) inthe flow phase. Binding is detected by a change in the SPR readout as described elsewhere (Gesellchen et al.,2005). B RI monomer (5 nM) was preincubated with cAMP concentrations ranging from 10 pM to 100 nM andthe binding to a high density 8AHA-cAMP sensor surface (3mM) was followed. The R subunit was injected for150 seconds on a Biacore 3000 instrument. 15 seconds after the end of the injection the binding values wererecorded (open circle) and the SPR signal was plotted against the logarithm of the cAMP concentration (Inset).The curve was fitted according to a sigmoidal dose response model and an apparent EC50 value was calculated. CApparent EC50 values of different analogues displayed as a radar plot. EC50 values ranging from 10�10 to 10�6 M

are plotted in a logarithmic scale as indicated. For abbreviations of the analogues see reagents

D. Moll et al.

Biomolecular interaction analysis

D. Moll et al.

ITC detects biological interactionswithout any modificationof the interaction partners

Using ITC (isothermal titration calorimetry)the heat released (or absorbed) due to ligandbinding is directly determined yielding infor-mation about the interaction stoichiometry(n), equilibrium binding constant (KA), freeenergy (�G), enthalpy (�H) and entropy(�S). The experimental setup of an ITCexperiment requires some knowledge con-cerning the nature of the molecules of inter-est and their inherent binding behaviour.

Wisemann et al. (1989) have shown, thatthe product of the binding constant (KA), thestoichiometry (n) and the protein concentra-tion has to be in the range between 1 and1000, preferably between 10 and 100 whenperforming an direct ITC experiment. Apply-ing the information to the PKA model sys-tem, where cAMP binds with high affinity tothe R subunit, the optimal protein concen-tration for the assay has to be rather low(500 pM–250 nM), which means in turn, thatthe heat evolved during the reaction at suchlow protein concentration is well below thedetection limit (data not shown). Therefore aless potent activator of the PKA holoenzyme,cGMP, was used in the experiment shown inFig. 5. Assuming that each RI monomer canbind two molecules cGMP, several differentmodels were applied to fit the binding iso-therm. Only a model with a single set of iden-tical sites yielded a satisfying fit (Fig. 5B)and a KD value of 117 nM could be calculated.This value was in the same range comparedto the data derived from SPR measurements(250 nM, Fig. 4C). Furthermore, based on the

Gibbs-Helmholtz equation (�G¼�H–T�S),valuable information about the thermody-namics of the interaction can be obtained withITC. The binding of cGMP to RI monomerwas found to be exothermic, i.e. with an re-action enthalpy (�H) of �15.4 kcalmol�1

and an entropy (�S) of �20 calmol�1 K�1.This results in a calculated free enthalpy(�G) of �9.5 kcalmol�1.

cAMP analogue characterizationin living cells using BRET2

A genetically encoded sensor was developedto measure PKA subunit dynamics inresponse to cAMP binding in intact cells(Prinz et al., 2006). The sensor was utilizedto characterize the effects of cAMP ana-logues on the PKA holoenzyme complex ina cellular assay which was adapted to a micro-titerplate (96 well) format. Bioluminescenceresonance energy transfer occurs when en-ergy generated by substrate oxidation byRluc is transferred to a GFP2 that is in closeproximity (1–10 nm). The BRET signal isdetermined by measuring the ratio of green(acceptor, 515 nm) over blue (donor, 410 nm)light. In the PKA holoenzyme, resonanceenergy transfer from the luciferase tagged Rsubunit to the GFP2 tagged C subunit occurs.However, when cAMP or PKA agonists bindto the holoenzyme, the BRET2 signal de-creases in response to increasing ligand con-centrations inside the cell. Here, the sensorwas used to analyze the efficacy of a mem-brane permeable PKA agonist, 8Br-cAMP-AM. In detail, COS-7 cells were seeded ina 96 well microtiterplate and co-transfectedwith plasmid DNAs coding for PKA type II

1Fig. 5. Analysis of cGMP binding to RI monomer employing ITC. A Titration of RI monomer (5 mM) with5.1mM cGMP (5 ml injection steps), raw data. The rate of heat release is plotted as a function of time. Inset:Schematic illustration of an ITC sample cell filled with R subunit (dark grey symbol) and injected cGMP (blackdots). As the two interacting partners form a complex, the released heat (arrows) is detected. For simplificationthe rotating syringe is not pictured. B Plot of heat exchange per mole of injectant (integrated areas under therespective peaks in A) vs. molar ratio of protein after blank subtraction (binding isotherm). The best least squaresfit applying a model where two cGMP per RI monomer are bound (one binding site model) was performed using

the software MicroCal Origin

Biomolecular interaction analysis

sensor. 48 hours after transfection, cells werewashed in D-PBS, and were incubated (15minutes) with various concentrations of8Br-cAMP-AM, ranging from 0 to 500 mM.

After addition of the luciferase substrateDeepblueCTM, bioluminescence generated bythe luciferase as well as GFP2-fluorescencewas quantified using a multi label reader.

Fig. 6. Sensing cAMP in living cells employing BRET2. A Schematic illustration of bioluminescence reso-nance energy transfer (BRET2). Luminescence is generated upon substrate (S) oxidation by Renilla reniformisluciferase (Rluc) and resonance energy is transferred to GFP2 in close proximity (1–10 nm) causing fluores-cence. The interaction of R subunit (dark grey symbol) and C subunit (light grey symbol) can thus bemonitored, for details see materials and methods. The BRET2-signal decreases when the holoenzyme dissoci-ates in response to rising concentrations of intracellular cAMP or to addition of cAMP analogues to the cells. BCells co-transfected with GFP-C and RII

a -Rluc were exposed to increasing concentrations of the membranepermeable PKA agonist 8Br-cAMP-AM. After addition of the luciferase substrate DeepBlueCTM, light emis-sion of the donor and the acceptor were determined and BRET2 ratios were calculated. Data are means� SD oftwo independent experimental setups, performed with n¼ 6 replicates at each concentration. A control value(line) was determined with cells co-expressing GFP2 and Rluc proteins solely and represents an average of ten

independent experiments

D. Moll et al.

Figure 6 shows a dose dependent reductionin BRET2 signal in cells expressing PKAtype II holoenzyme when treated with up to500 mM analogue (maximum stimulation).From these data, a half maximal effective con-centration of about 13 mM for 8Br-cAMP-AM was calculated. The AM substitutionleads to an increase in efficacy by two ordersof magnitude in comparison to the parentcompound 8Br-cAMP, which has an EC50

value in the millimolar range (about 1.5–2mM, data not shown).

Discussion

The classical proteomics approach comprisesthe differential description of protein ex-pression patterns using established methodslike 2D polyacrylamide gel electrophoresiscombined with mass spectrometry. This strat-egy was expanded by functional proteomicsanalyzing the distinct interaction patterns ofintracellular protein assemblies, and by che-mical proteomics trying to construct syntheticentities interfering with previously definedinteractions. Recently, interaction networkshave been analyzed across an entire genomeusing for example the tap technology (tan-dem affinity purification (Gavin et al., 2002))or a genome wide yeast two hybrid approach(Ho et al., 2002).

Those methods generate a vast amount ofbinding data and require bioinformatics to in-terconnect the molecular complexes and path-ways within a cellular network, for examplethe database BIND (Bader et al., 2001). How-ever, these data only provide a qualitativedescription of a given interaction network.The pharmaceutical industry and biotech re-search have a strong interest in quantifyingthe number and strength of intracellular inter-actions. Furthermore, information about post-translational modifications and subcellularlocalization have to be included to compre-hensively describe cellular networks.

Biological binding events cover a broadrange of affinities from low picomolar (e.g.

streptavidin-biotin interaction (Green, 1990))via the nanomolar range for ligand receptor-or antibody antigen-binding, to the micromo-lar range for signalling modules like SH3- orSH2-domains. However, even affinities in themillimolar range may be of physiologicalsignificance considering high local proteinconcentrations mediated by compartmentali-zation. The affinity of a given interaction pairmay vary by several orders of magnitudedepending on the situation in the cell. Forexample, the R and C subunits of PKA bindwith an affinity of 1 pM in the absence ofcAMP which is shifted to 1 mM (factor106!) in the presence of cAMP (Kopperudet al., 2002). Therefore, a high demand fortechnologies exists capable of analyzing thecomplete bandwidth of biological affinities.Biomolecular interaction analysis (BIA) com-bines a variety of assay designs trying to gen-erate highly accurate binding data. In thisarticle established and novel BIA methodsare compared side by side using PKA as amodel system. AlphaScreen provides a highthroughput capable, homogenous, bead basedassay platform with luminescence readout.FP, another high throughput method, requiresfluorescent labelling of one interaction part-ner and is used for rapid determinationof equilibrium binding constants. SPR usingBiacore instruments was chosen as a highlyreliable method to monitor association anddissociation kinetics separately. The onlytruly label free method employed here isITC generating additional information aboutthe thermodynamics of a given interaction.Finally, a cell based assay, BRET2, was usedto quantify protein–protein interaction inresponse to low molecular weight ligands inliving cells. Since every method has certainadvantages and inherent drawbacks the fol-lowing discussion provides a brief assess-ment of each technology.

AlphaScreen

AlphaScreen is a robust method for thedetection of biological interactions. Because

Biomolecular interaction analysis

the measurement is time-resolved and lightemission occurs at a higher wavelength thanexcitation, the background is very low, re-sulting in a good signal-to-noise ratio. Asopposed to other methods presented here,AlphaScreen is usually not feasible for thedetermination of absolute binding constants,since the concentrations of interaction part-ners on the surfaces of the beads can not becontrolled very well. On the other hand, highlocal concentrations of biomolecules andconcurrent avidity effects allow for the detec-tion of low affinity interactions. This resultsin a high dynamic range from picomolar tomillimolar concentrations. The method yieldsreproducible EC50 values and permits aneffective screening of compounds that affectbinding. This is also reflected in the PKAmodel system tested here (Fig. 2), wherethe relative potency of a cyclic AMP analogue(Sp8Br-cAMPS) is determined correctly com-pared to cAMP (Schaap et al., 1993).

In principle, any type of interaction canbe measured with AlphaScreen provided thatthe interaction partners can be immobilizedto either donor or to acceptor beads. In con-trast to direct coupling of biomolecules to thebead surface one or both interaction partnersare commonly captured on antibody-coatedbeads. Biotinylation of one interaction part-ner, that can subsequently bind to streptavi-din coated beads, is another option. Oneshould be aware of the fact that these in-direct methods introduce new equilibria intothe interaction, which might influence themeasurement.

FP

Fluorescence polarization (FP) can be usedto directly determine EC50 values betweenfluorescently tagged molecule and its bind-ing partner ideally with a relative molecularweight ratio of 1:50 or higher (Burke et al.,2003). Binding can then be monitored simplyby reading the polarization value of the fluo-rescent molecule after it binds to a largermolecule. Developing competitive binding

assays, which can be used to quantify unla-belled competitors in solution, is straightfor-ward (Burke et al., 2003).

The PKA model system can be usedafter chemical modification of cyclic AMPwith a fluorophore. In this study, Fluoresceinwas attached to the C8 position of theadenine moiety. Fluorescein is rather bigcompared to cAMP (cAMP: 329Da; 8Fluo-cAMP, 816Da), thus adding the fluorophoremight significantly affect the interactionwith the RI monomer (33.000Da). There-fore the binding behaviour of fluorescentlylabelled cAMP was quantified in independentassays. Interestingly, employing the direct FPassay, our results indicate that attaching afluorophore at the C8 position does not affectthe binding affinity to the RI monomer (EC50

1.7 nM). This was verified by 3H-cAMPbinding assays (D. Moll, N. C. G. Burghardt,unpublished results, 1.1 nM) and by SPRexperiments (see Fig. 4C, 1.7 nM). Further-more, based on a phosphotransferase activityassay (Cook et al., 1982), the activation con-stant of the PKA holoenzyme was not altered(D. Moll, unpublished results). Full lengthRI and RII subunits of PKA purified frombovine skeletal muscle and heart, respec-tively, confirmed that 8Fluo-cAMP andcAMP bind comparably well (Mucignat-Caretta and Caretta, 1997).

EC50 values for the analogues 2AHA-cAMP and 6AH-cAMP are in perfect agree-ment with the SPR competition results(2AHA-cAMP: SPR 33 nM, FP 27.8 nM;6AH-cAMP: 1.4 nM for both SPR and FP(Figs. 3B and 4C)). 8AHA-cAMP and cAMPshow similar affinities in both assays withslightly higher EC50 values generated withFP (cAMP: SPR 1.7 nM, FP 4 nM; 8AHA-cAMP: SPR 1.5 nM, FP 5.4 nM; Figs. 3Band 4C).

FP assays have been developed for near-ly all major classes of drug targets, includ-ing GPCR’s (G-Protein coupled receptors),kinases, phosphatases, proteases and nuclearreceptors (Burke et al., 2003).

D. Moll et al.

SPR

SPR measurements have been utilized in thedrug development process and can also beused in functional proteomics in order tokinetically characterize binding events withincellular interaction networks. Typical ap-plications are target characterization andvalidation of pre-selected binders in second-ary screens. Furthermore, lead optimization,QSAR (quantitative structure activity rela-tionship) and early ADME studies (adsorp-tion, distribution, metabolism, excretion) canbe performed. Employing SPR almost everybiomolecular interaction can be measured. Ina Biacore system not only equilibrium bind-ing data but also rate constants for the asso-ciation and dissociation can be determinedallowing the characterization of a wide rangeof affinities with high reproducibility, highsensitivity and low sample consumption.

Potential limitations of SPR biosensorslie in the detection principle: a mass changeclose to the sensor surface is converted intoan optical signal, meaning that a small massincrease results in an accordingly small sig-nal. Consequently the immobilization of thelow molecular weight interaction partner ap-pears to be favourable. Immobilization ofsmall ligands, however, often requires theirderivatization which may affect the function-ality of a given ligand.

A general problem of solid phase assays isa reduction in degrees of freedom upon immo-bilization, which can severely influence oreven prevent binding due to steric hindrance.

Although SPR based biosensors are com-monly used in the direct assay format, alter-natively the binding of low molecular weightsubstances can be assessed in solution or sur-face competition assays as shown here alsofor AlphaScreen and FP. In the solution com-petition setup, the competitor molecule inter-feres with the binding of the analyte to theimmobilized ligand, whereas the surface com-petition experiment the molecule of interestcompetes with the analyte for the same bind-

ing site. In work presented here, the relativebinding affinities (EC50 values Fig. 4C) ofseveral cyclic nucleotide derivatives contain-ing different functional modifications weretested in form of a solution competition as-say. This also avoided problems arising fromsteric hindrance and from mass transfer lim-itations (Glaser, 1993). With this approach itwas possible to cover affinities of severalorders of magnitude. Still, depending on theposition where the cyclic nucleotide wasimmobilized to the sensor surface, stericeffects had to be considered. However, con-trol experiments demonstrated that immobi-lizing cyclic AMP via aminohexyl linkers atthe two, six or eight position of the adeninering did not influence the EC50 values (datanot shown).

ITC

Isothermal titration calorimetry (ITC) hasbeen applied to a variety of biochemicalinteractions. In contrast to all techniques dis-cussed so far, no chemical modification ofthe target molecule is needed for immobi-lization or labelling. Besides equilibriumbinding data, ITC delivers thermodynamicparameters such as enthalpy (�H), entropy(�S) and the stoichiometry (n). However,ITC may require rather large amounts of theinteraction partners especially of the mole-cule in the syringe, which is usually used in7n-fold molar excess over the molecule inthe cell. The interaction between RI mono-mer and the effector cGMP was shown tobe exothermic. This allowed to measurecGMP=R subunit interaction at relativelylow protein concentration using less than6 mM RI monomer.

(Gorshkova et al., 1995; Lin andLee, 2002)could demonstrate that cGMP binding to thetwo nucleotide binding sites in the E. colicatabolite activator protein (CAP) is alsoexothermic (�H¼�1.7, �2.7 kcalmol�1),however, 10-fold less compare to RI mo-nomer (�H¼�15.4 kcalmol�1).Accordinglya much higher concentration of CAP pro-

Biomolecular interaction analysis

tein (50–400mM)was required for ITC experi-ments. Interestingly the cGMP binding to CAPis entropically favoured with a positive �Saround þ13 calmol�1 K�1 (Gorshkova et al.,1995; Lin and Lee, 2002) in contrast to cGMPbinding to RI monomer with an unfavourablenegative �S¼�21 calmol�1 K�1.

BRET2

With the use of GFP as well as luciferasesfrom several organisms as genetically en-coded sensors in living cells, two main reso-nance energy based methods (FRET andBRET) for the investigation of protein–pro-tein interactions have emerged. For a detailedcomparison refer to (Boute et al., 2002; Xuet al., 2003). In contrast to the traditionalyeast two hybrid assay, where protein–pro-tein interactions are commonly performedin the nucleus, FRET and BRET based bio-molecular interactions can be monitored indefined sub-cellular compartments. In ourstudy we chose a BRET2 based reporter sys-tem to quantify holoenzyme dissociationin vivo, because BRET2 represents a ratio-metric assay. BRET2 can be used with a de-fined population of cells as well as in celllysates (not shown). Compounds influencingholoenzyme formation were investigated andtested for efficacy and bioavailability, com-plementing established in vitro assays. Withthis BRET2 reporter system, we investigatedthe effect of the membrane permeable ago-nist 8Br-cAMP-AM on intracellular PKAtype II holoenzyme, which results in a dosedependent BRET2 signal reduction follow-ing the dissociation of the holoenzyme. 8Br-cAMP-AM is far more potent (about 50times) compared to its parent compound 8Br-cAMP (data not shown), with an EC50 valueof 13 mM (Fig. 6). The data are in good agree-ment with earlier results testing the potencyof acetoxymethyl esters of Bt2-cAMP usingphysiological tests and microinjected PKAtype II FRET sensors (Schultz et al., 1993).EC50 data in the same range were reported

in anti-proliferative studies with several acet-oxymethylesters of cAMP (Bartsch et al.,2003) and in studies on the anti-apoptoticproperties of 8Br-cAMP-AM in neutrophilschallenged with TNF-a (Krakstad et al.,2004).

BRET based assays can be easily up-graded to medium high throughput formatunder standardized conditions. With thedevelopment of novel microscopic and im-aging techniques, resonance energy basedmethods could be useful for mapping signal-ling events in cells and even living organisms(Tsien, 2003).

Acknowledgements

We thank Angelika Wattrodt, Mandy Diskar andMichael Krieg (University of Kassel, Germany) for su-perb technical help andDr. Frank Schwede,Dr. GottfriedGenieser (Biolog, Bremen) for cyclic nucleotide ana-logues, respectively. We acknowledge Dr. ClaudiaHahnefeld and Dr. Christian Hammann (University ofKassel, Germany) for helpful discussions regardingSPR and ITC measurements. This work was supportedby grants of the Deutsche Forschungsgemeinschaft(DFG, He1818=4) and the European Commission(EU-RTD QLK3-CT-2002-02149) and by the GermanMinistry of Education and Research (BMBF PPO-S22T02) to F. W. Herberg.

References

Arnsten AF, Ramos BP, et al. (2005) Protein kinase Aas a therapeutic target for memory disorders:rationale and challenges. Trends Mol Med 11:121–128

Bader GD, Donaldson I, et al. (2001) BIND-TheBiomolecular Interaction Network Database.Nucleic Acids Res 29: 242–245

Bartsch M, Zorn-Kruppa M, et al. (2003) Bioac-tivatable, membrane-permeant analogs of cyclicnucleotides as biological tools for growth controlof C6 glioma cells. Biol Chem 384: 1321–1326

Blume-Jensen P, Hunter T (2001) Oncogenic kinasesignalling. Nature 411: 355–365

Boute N, Jockers R, et al. (2002) The use of resonanceenergy transfer in high-throughput screening:BRET versus FRET. Trends Pharmacol Sci 23:351–354

Buechler YJ, Herberg FW, et al. (1993) Regulation-defective mutants of type I cAMP-dependent pro-

D. Moll et al.

tein kinase. Consequences of replacing arginine 94and arginine 95. J Biol Chem 268: 16495–16503

Burke TJ, Loniello KR, et al. (2003) Development andapplication of fluorescence polarization assays indrug discovery. Comb Chem High ThroughputScreen 6: 183–194

Cook PF, Neville ME Jr, et al. (1982) Adenosine cyclic30,50-monophosphate dependent protein kinase:kinetic mechanism for the bovine skeletal musclecatalytic subunit. Biochemistry 21: 5794–5799

Fabbro D, Ruetz S, et al. (2002) Protein kinases astargets for anticancer agents: from inhibitors touseful drugs. Pharmacol Ther 93: 79–98

Gavin AC, Bosche M, et al. (2002) Functional orga-nization of the yeast proteome by systematic ana-lysis of protein complexes. Nature 415: 141–147

Gesellchen F, Zimmermann B, et al. (2005) Directoptical detection of protein-ligand interactions.Methods Mol Biol 305: 17–46

Glaser RW (1993) Antigen-antibody binding and masstransport by convection and diffusion to a surface: atwo-dimensional computer model of binding anddissociation kinetics. Anal Biochem 213: 152–161

Gorshkova I, Moore JL, et al. (1995) Thermodynamicsof cyclic nucleotide binding to the cAMP receptorprotein and its T127L mutant. J Biol Chem 270:21679–21683

Graves PR, Haystead TA (2002) Molecular biologist’sguide to proteomics. Microbiol Mol Biol Rev 66:39–63

Green NM (1990) Avidin and streptavidin. MethodsEnzymol 184: 51–67

Hagiwara M, Brindle P, et al. (1993) Coupling ofhormonal stimulation and transcription via thecyclic AMP-responsive factor CREB is rate limitedby nuclear entry of protein kinase A. Mol Cell Biol13: 4852–4859

Herberg FW, Bell SM, et al. (1993) Expression of thecatalytic subunit of cAMP-dependent proteinkinase in Escherichia coli: multiple isozymesreflect different phosphorylation states. ProteinEng 6: 771–777

Herberg FW, Dostmann WR, et al. (1994) Crosstalkbetween domains in the regulatory subunit ofcAMP-dependent protein kinase: influence ofamino terminus on cAMP binding and holoenzymeformation. Biochemistry 33: 7485–7494

Ho Y, Gruhler A, et al. (2002) Systematic identificationof protein complexes in Saccharomyces cerevisiaeby mass spectrometry. Nature 415: 180–183

Jameson DM, Croney JC, et al. (2003) Fluorescence:basic concepts, practical aspects, and some anec-dotes. Methods Enzymol 360: 1–43

Jameson DM, Mocz G (2005) Fluorescence polariza-tion=anisotropy approaches to study protein-ligand

interactions: effects of errors and uncertainties. In:Nienhaus GU (ed) Methods in Molecular Biology:Protein Ligand Interactions: Methods and Applica-tions. Humana Press Totowa, pp 301–322

Kandel ER (2001) The molecular biology of memorystorage: a dialogue between genes and synapses.Science 294: 1030–1038

Knighton DR, Zheng JH, et al. (1991) Structure of apeptide inhibitor bound to the catalytic subunit ofcyclic adenosine monophosphate-dependent pro-tein kinase. Science 253: 414–420

Kopperud R, Christensen AE, et al. (2002) Formationof inactive cAMP-saturated holoenzyme of cAMP-dependent protein kinase under physiological con-ditions. J Biol Chem 277: 13443–13448

KrakstadC,ChristensenAE, et al. (2004) cAMPprotectsneutrophils against TNF-alpha-induced apoptosisby activation of cAMP-dependent protein kinase,independently of exchange protein directly activatedby cAMP (Epac). J Leukoc Biol 76: 641–647

Lakowicz JR (1999) Fluorescence Anisotropy. Princi-ples of fluorescence Spectroscopy. Kluwer acade-mic=Plenum Publishers, New York, pp 291–319

Lin SH, Lee JC (2002) Communications between thehigh-affinity cyclic nucleotide binding sites in E.coli cyclic AMP receptor protein: effect of singlesite mutations. Biochemistry 41: 11857–11867

Manning G, Whyte DB, et al. (2002) The proteinkinase complement of the human genome. Science298: 1912–1934

Mucignat-Caretta C, Caretta A (1997) Binding of twofluorescent cAMP analogues to type I and II reg-ulatory subunits of cAMP-dependent proteinkinases. Biochim Biophys Acta 1357: 81–90

Olsen SR, Uhler MD (1989) Affinity purification of theC alpha and C beta isoforms of the catalytic subunitof cAMP-dependent protein kinase. J Biol Chem264: 18662–18666

Prinz A, Diskar M, et al. (2006) Detection of humanprotein kinase A type I and type II subunit inter-action in intact cells using bioluminescence reso-nance energy transfer. Cellular Signalling, DOI:10.1016=j.cellsig.2006.01.013

Sachsenmaier C (2001) Targeting protein kinases fortumor therapy. Onkologie 24: 346–355

Sambrook J, Russel DW, et al. (2001) MolecularCloning – A laboratory manual. Cold Spring HarborLaboratory Press, New York, pp 15.36–15.39

Schaap P, van Ments-Cohen M, et al. (1993) Cell-permeable non-hydrolyzable cAMP derivatives astools for analysis of signaling pathways controllinggene regulation in Dictyostelium. J Biol Chem 268:6323–6331

Schultz C, Vajanaphanich M, et al. (1993) Acetoxy-methyl esters of phosphates, enhancement of the

Biomolecular interaction analysis

permeability and potency of cAMP. J Biol Chem268: 6316–6322

Shaywitz AJ, Dove SL, et al. (2002) Analysis ofphosphorylation-dependent protein–protein inter-actions using a bacterial two-hybrid system. SciSTKE 2002: PL11

Slice LW, Taylor SS (1989) Expression of thecatalytic subunit of cAMP-dependent proteinkinase in Escherichia coli. J Biol Chem 264:20940–20946

Tasken K, Aandahl EM (2004) Localized effects ofcAMPmediated by distinct routes of protein kinaseA. Physiol Rev 84: 137–167

Taylor SS (1989) cAMP-dependent protein kinase.Model for an enzyme family. J Biol Chem 264:8443–8446

Taylor SS, Yang J, et al. (2004) PKA: a portrait ofprotein kinase dynamics. Biochim Biophys Acta1697: 259–269

Tsien RY (2003) Imagining imaging’s future. Nat RevMol Cell Biol [Suppl]: SS16–SS21

Valeur B, Brochon JC eds. (2001) New Trends inFluorescence Spectroscopy – Applications to Che-mical and Life Sciences. Springer Series on Fluor-escence – Methods and Applications. SpringerVerlag, Heidelberg

Wiseman T, Williston S, et al. (1989) Rapid measure-ment of binding constants and heats of bindingusing a new titration calorimeter. Anal Biochem179: 131–137

Xu Y, Kanauchi A, et al. (2003) Bioluminescenceresonance energy transfer: monitoring protein–pro-tein interactions in living cells. Methods Enzymol360: 289–301

Yaqub S, Abrahamsen H, et al. (2003) Activation ofC-terminal Src kinase (Csk) by phosphorylationat serine-364 depends on the Csk-Src homology3 domain. Biochem J 372: 271–278

Author’s address: Friedrich W. Herberg, Univer2-sity of Kassel, Heinrich-Plett-Str. 40, 34132 Kassel,Germany, e-mail: herberg@uni-kassel.de

D. Moll et al.: Biomolecular interaction analysis