Analysis of virtual two-dimensional gels based upon affinity capillary electrophoresis hyphenated to ion trap-mass spectrometry

Nadine Machour1

John Place1

François Tron1

Roland Charlionet1

Laurent Mouchard2

Christophe Morin3

Annie Desbène3

Paul-Louis Desbène3

1INSERM U519,Faculté Mixte de Médecine –Pharmacie et IFRMP,Rouen, France

2Atelier de BioInformatiqueStatistique et Sociolinguistique,Faculté des Sciences,Mont-Saint-Aignan, France

3Laboratoire d’Analyse desSystèmes OrganiquesComplexes UPRES 3233 (SMS)IRCOF et IFRMP,Evreux, France

Analysis of virtual two-dimensional gels based uponaffinity capillary electrophoresis hyphenated to iontrap-mass spectrometry

Affinity capillary electrophoresis (ACE) is a robust tool for the study of noncovalentbiomolecular interactions and to determine the binding constants. It is advantageousdue to the speed of analysis, the high and reproducible separation efficiencies, the lowconsumption of analytes, the ability to study several interactions at the same time, andto cover a wide range of affinity. The use of an ion trap-mass spectrometer as a sensi-tive and specific detector, coupled on-line with a classical UV detector, permitsextracting simultaneously the electropherograms corresponding to each ionic species.The mass spectra, acquired by scanning the results of a first separation due to ACE,were assimilated into a virtual two-dimensional (2-D) gel. We developed a softwareapplication, which was designed to create and analyze these virtual 2-D gels. Thevalidity of this new analytical tool for probing biomolecular interactions has beendemonstrated on mixtures of antibiotics of the vancomycin group and several dipep-tide substrates. Using the dynamic equilibrium affinity electrophoresis approach, wehave shown that molecular components interacting with a low affinity are easily locatedon the virtual 2-D gels, and that binding constants and stoichiometry of the interactionscan be assessed. As the binding constants derived from ACE-electrospray ionization-mass spectrometry (ESI-MS) are unreliable, they must only be determined with the UVdetector.

Keywords: Affinity capillary electrophoresis / Functional proteomics / Ion trap-mass spectrometry /Virtual two-dimensional gels DOI 10.1002/elps.200410213

1 Introduction

Interacting proteins in metabolic and signalling pathwaysas well as in complex molecular machines control the lifeof biological cells. Identification of specific binding andestimation of binding constants are essential for translat-ing the vast amount of data that now exist on genomesand proteomes into an understanding of how livingorganisms work. Functional proteomics research requiresthe development of novel methodologies to characterizeprotein interactions in complex mixtures. In recent years,affinity capillary electrophoresis (ACE) has been intro-duced as an effective and versatile tool for studying bio-molecular noncovalent interactions and determiningbinding and dissociation constants [1–3]. ACE is advan-tageous due to the speed of analysis, the high and repro-ducible separation efficiencies, and the low consumption

of analytes. The theory of ACE borrows its concepts fromseparation sciences and biochemistry: it rests mainlyupon the application of the mass balance law to molecu-lar equilibria. There are numerous ACE modes, which canbe used for a wide variety of applications and which cancover a wide range of affinity interactions. Ligands mayinteract with target molecules (receptors) before and/orduring electrophoresis [4]. In the latter case, the receptorsare added to the running buffer or immobilized in thecapillary and the ligands move through a constant con-centration of receptors. This approach belongs to dy-namic equilibrium affinity electrophoresis, also calledmobility shift ACE. It is especially well-suited for bindingconstant estimates corresponding to low-to-intermediateaffinity interactions that are the object of the work pre-sented here. When the interaction kinetics are suitable,i.e., when the binding equilibrium is fast compared to theelectrophoresis migration rate, the free and bonded mo-lecular species migrate as one band in the capillary. In thistype of ACE, the binding constant is derived by analyzingthe gradual peak migration shifts of ligands as a functionof the receptor concentration in the separation buffer.Several equations linking the mobility shift with thereceptor concentration and the association constant have

Correspondence: Dr. Roland Charlionet, INSERM U519, FacultéMixte de Médecine – Pharmacie, 22 boulevard Gambetta,F-76183 Rouen cedex, FranceE-mail: [email protected]: 133-2-3514-8541

Abbreviation: ACE, affinity capillary electrophoresis

1466 Electrophoresis 2005, 26, 1466–1475

2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Electrophoresis 2005, 26, 1466–1475 ACE hyphenated to ion trap-MS 1467

been developed [3–9]. They are derived from the massbalance law applied to the equilibrium between ligand L,receptor R, and complex RL. The equilibrium associationconstant K is:

K = [RL]/[R]? [L] (1)

The net electrophoretic mobility of the ligand resultingfrom its free and complexed form can be introduced [10]and used to rearrange Eq. (1) as following:

Dm = Dmmax?K? [R]/(11K? [R]) (2)

Dm is the shift of the apparent electrophoretic mobility ofthe ligand in the absence and in the presence of thereceptor; Dmmax stands for the shift of the apparent elec-trophoretic mobility between the free ligand and theligand-receptor complex. Experiments based on Eq. (2)can be designed to obtain the equilibrium constants ofinteractions.

The interest of using a mass spectrometer as a sensitiveand specific detector of separations realized with ACEhas been shown in some studies [11, 12]. The on-linehyphenation of CE to MS detection allows obtainingstructural elucidation to some extents [13]. Indeed, fewmeasurements in biochemistry are as fundamental as theaccurate determination of the mass of a component [14].It is the key tool to facilitate identification of proteins,characterization of drugs and their metabolites, screeningof biochemical constituents present in biological fluids,cells or tissues, and elucidation of physiological mechan-isms. The on-line coupling of ACE with MS provides themeans to extract simultaneously the electropherogramscorresponding to each ionic species in a mixture of bio-logical compounds and may be used to determine multi-ple binding constants [11, 12]. The idea of assimilating themass spectra, acquired by scanning the results of a firstseparation technique, into a “virtual” 2-D gel has recentlyemerged [15–18]. In the first place it concerned the use ofMALDI-TOF-MS for the detection of intact proteins di-rectly from the surface of an immobilized pH gradient gel.We propose here a new methodology: the analysis of vir-tual 2-D gels, which have been realized by on-line cou-pling of ion trap-MS with ACE. Mass spectra wereacquired by analyzing the outflow of the capillary andwere assimilated into a virtual 2-D gel by an appropriatedata processing. This virtual 2-D gel is similar to a classi-cal 2-D gel, except that for the abscissa the pI is replacedby the migration time and that for the ordinate the molec-ular weight is acquired by ion trap rather than by SDS-PAGE. We elaborated a software application to analyzethese virtual 2-D gels. It enables us to characterize thechanges occurring to the ligands expression associatedwith the concentration of their target molecules. Detec-tion of virtual 2-D gel spots and matching of virtual 2-D

gels have been implemented in the software and we ana-lyzed the matching results in order to give estimates of thebinding constants. Thus, the differential analysis of virtual2-D gels gives access to the identification, quantitation,and characterization of the binding interactions occurringin biological fluids.

In order to establish the validity of our novel analytical toolfor probing biomolecular interactions, we have beenusing not only the classical model system vancomycin/D-Ala-D-Ala [11, 12, 19–23], one of the most extensivelystudied interactions in biomolecular recognition [24], butalso other systems based on teicoplanin and ristocetin.These glycopeptide antibiotics of the vancomycin groupplay an important role against Gram-positive infectionsgiven that many bacteria have become resistant to peni-cillin and a range of other antibiotics. It is by now well-established that the antibiotic activity of vancomycin is adirect result of noncovalent interactions between the drugand the peptidoglycan precursors of the bacterial cellwall. More specifically, vancomycin targets the (D-Ala-D-Ala) terminus of the peptidoglycan. During the growth ofnormal cell wall, specific bacterial enzymes use thesetermini to form new cross-links in the peptidoglycan.Antibiotics of the vancomycin group can inhibit the bac-terial cell wall growth since they bind competitively to thesame carboxy-terminal sequence, ultimately leading tobacterial cell death. In other respects, antibiotics of thevancomycin group exhibit complex chemical structures.The peptide backbone accommodates different carbo-hydrate moieties as well as several chemical derivativesand presents numerous stereogenic centers. It has alsobeen established that the formation of dimers of theseglycopeptide antibiotics affects their binding efficacy withtheir ligands [25]. Thus, using different mixtures of van-comycin related antibiotics, we intended to demonstratethat analyzing virtual 2-D gels, built from on-line couplingof ion trap-MS with ACE, can be an effective means tocharacterize interacting molecules in a mixture of biolog-ical compounds and to measure the set of their biomole-cular interactions.

2 Materials and methods

2.1 Chemicals

All solutions were prepared by using 18 MO water pro-duced by means of an AlphaQ purification system (Milli-pore, Bedford, MA, USA). Acetic acid and ammoniumacetate used for the preparation of the running buffercame from Acros (Acros, Noisy-Le-Grand, France).DMSO used as EOF marker in UV detection was alsoprovided by Acros. Vancomycin hydrochloride (from


CE

and

CE

C

1468 N. Machour et al. Electrophoresis 2005, 26, 1466–1475

Streptomyces orientalis), ristomycin monosulfate, N,N-diacetyl-Lys-D-Ala-D-Ala (peptide 1), N,N-diacetyl-Lys-D-Ala-D-lactic acid (peptide 2), N-acetyl-Ala-Ala-Ala (pep-tide 3), and N-acetyl-D-Ala-D-Ala (peptide 4) were pur-chased from Sigma (Saint Quentin Fallavier, France).Interchim (Montluçon, France) provided teicoplanin hy-drochloride. Methanol used in the preparation of thesheath liquid was of RS-HPLC purity and was obtainedfrom Carlo Erba (Val de Reuil, France).

2.2 Affinity capillary electrophoresis

The ACE experiments were carried out on a P/ACE 2100capillary electrophoresis (Beckman-Coulter, Fullerton,CA, USA), fitted with UV detection, and modified forhyphenation with MS detection as previously described[13]. The analytes were monitored at 214 nm. Acquisitionswere performed by means of P/ACE 2000 softwareVersion 2.0 (Beckman-Coulter). Typically, the sampleswere injected in hydrodynamic mode at 0.5 psi, i.e.,3.4 kPa for 5 s. The operating temperature was 257C andthe applied voltage was 125 kV. The electrophoreticexperiments were done with untreated fused-silica capil-laries purchased form Thermo-Ionic France (Courtaboeuf,France). The capillaries dimensions were 57 cm (50 cmeffective length)650 mm ID6375 mm OD when UVdetection was used alone or 80 cm (20 cm effective lengthfor UV detection)650 mm ID6375 mm OD when UV andMS detections were simultaneously performed. The cap-illary tips were cut by means of a shortix capillary columncutter (Agilent Technologies France, Massy, France). ThepH of the running electrolytes was measured at 257C witha model f pH meter (Beckman-Coulter). The electrolyteswere systematically degassed by sonication by means ofan Ultrasonic Cleaner model 2510 (Branson Ultrasonics,Danbury, CT, USA). New capillaries were successivelyflushed with 0.5 M sodium hydroxide for 45 min and thenwith water (15 min). Between two consecutive injectionsthe rinsing procedure was 5 min with water, 10 min with0.1 M NaOH, 5 min with water, and finally 5 min with therunning electrolyte (18.561023 M ammonium acetate,pH 5.1) containing the studied peptide at the requiredconcentration. The injected solutions contained vanco-mycin, ristocetin, or teicoplanin (or a mixture of theseones) in the running electrolyte. We systematically per-formed a co-injection of 15.561023 M DMSO for 5 s inorder to mark the EOF.

2.3 Ion trap-mass spectrometry

The ACE-ESI-MS experiments were performed using anESQUIRE-LC ion trap mass spectrometer (Bruker, Wis-sembourg, France) hyphenated with a P/ACE 2100 capil-

lary electrophoresis instrument by means of an HP3D CE/ESI-MS interface (Agilent Technologies, Massy, France).For this coupling, it should be noted that the heightdifference between the two tips of the capillary was lessthan 1 mm to prevent a siphoning effect [26]. Moreover,the position of the capillary within the nebulizing needlewas adjusted by means of a micrometry screw to obtain asatisfactory reproducibility as regards migration times,efficiencies, and MS sensitivity (required accuracy, 40 mm)[26]. The sheath liquid was H2O/methanol/acetic acid(10/10/2.5 v/v/v) at a flow rate of 3 mL/min, adapted toobtain a satisfactory mass detection. It was delivered bymeans of a single-syringe infusion pump (Cole-Parmer,Vernon Hills, IL, USA) equipped with 250 mL Hamiltongastight syringes (Hamilton, Bonaduz, Switzerland).Nitrogen was used as nebulizing and drying gas. It washeated to 2007C and its flow rate was 210 L/h. This gaswas generated by a Nitrox generator model UHPLCMS 18(Domnik Hunter Limited, Durham, UK). The air provided tothis generator was produced by a JUN’AIR compressormodel 2000-40B (Nørresundby, Denmark) equipped witha Cirrus refrigeration dryer model CGB 0025 purchasedfrom Domnik Hunter. The spectra were recorded in thepositive ionization mode. ESI was performed at thefollowing cap and end plate voltages: Ucap = 23.5 kV; Uend

plate = 23.0 kV. In order to optimize the experimentalconditions of detection in the electrospray ion trap-massspectrometer, preliminary experiments were performedby flow injection analyses.

2.4 Virtual 2-D gel

We developed a dedicated software that performs thevarious steps necessary to efficiently process the dataissued from signals transmitted by the ion trap: installingthe data processing structures, characterizing the spotsin virtual gels, carrying out the differential analysis, anddetermining the binding constants. The ion trap dataprocessing comes through the application program DataAnalysis (Bruker Daltonics). These data were exploiteddirectly with Gnuplot. A graphical interface combiningPerlTk and Gnuplot has been designed. The ionic cur-rents corresponding to the mass/charge ratios (m/z) ofinterest were extracted from the total ionic current. Peakdetection subroutines were applied to the data to detectany peak over the set threshold. The data were assem-bled into a single text file, which leads to obtaining theentire set of extracted ionic currents acquired from theMS scan of the electrophoresis run and consequently thevisualization of a 3-D plots (the ionic current as a functionof the migration time and of the m/z). From the 3-D plot, abidimensional representation of the electropherogramwas built: the virtual 2-D gel. As with classical 2-D gel,



each molecular species was characterized by its migra-tion time (on the abscissa), by its molecular weight (on theordinate), and by its ionic current on a colored scale linkedto the ionic intensity in the MS trap. The ligands, injectedinto the capillary, were detected in their free and bondedforms as peaks in the electropherogram and thereforeappeared as spots in the virtual 2-D gel. In our study,molecular species used as receptors were incorporatedinto the running buffer. They appeared in the virtual 2-Dgel as continuous lines parallel to the abscissa.

3 Results and discussion

3.1 Characterization of the interactingmolecular species

The electropherograms of MS-extracted ions are graphi-cal representations that give the ionic current intensityversus the migration time for each m/z ratio. So, studyingthe many interactions that occur in a complex mixture ofmolecular compounds requires a whole series of electro-pherograms. This fragmented perception does not facil-itate a clear understanding of the set of interactions. Forovercoming this difficulty, the whole series of the extrac-ted ion electropherograms can be gathered into a solerepresentation leading to a 3-D visualization of ACE anal-ysis. In such a 3-D plot, the evolution of the ionic current isshown as a function of the migration time and of m/z. The

3-D visualization of ACE analysis of a mixture constitutedof vancomycin, teicoplanin, and ristomycin with theirsusbstratum, N,N-diacetyl-Lys-D-Ala-D-Ala (peptide 1),which is present in the BGE, is reported in Fig. 1. BGEcontained 18.561023 M ammonium acetate and261025 M peptide 1. From the projection of the ionic cur-rent intensities in the plan defined by migration time(X-axis) and mass (Y-axis), we obtained a 2-D plot thatlooks like a classical 2-D gel. In this representation, spotsthat have parameters similar to those of the correspond-ing peaks reveal free and bonded ligands. The ordinate ofthe spot center corresponds to the m/z of the concernedmolecular species whereas its abscissa is the migrationtime of the peak apex visualized in the extracted ionelectropherogram for this m/z. The migration time corre-sponding to the lower and upper bounds of the peaks arethose of the spots. The peak width at half height is alsoindicated on the spots and a colored scale is used toindicate the MS signal intensity. As shown in Fig. 2, for agiven migration time, two spots can be visualized. Forexample, for a migration time equal to 10.4 min, one spotat m/z 725, corresponding to free vancomycin, and asecond one at m/z 911, for vancomycin complexed withpeptide 1, are detected. This is characteristic of low-af-finity interactions with a binding equilibrium fast com-pared to the electrophoresis migration. The peptide 1,which was introduced continuously in the BGE, appearsas a continuous line parallel to X-axis with the m/z of thisreceptor. The three studied antibiotics form complexes

Figure 1. 3-D representation ofthe interactions of the threeantibiotics related to vancomy-cin with peptide 1 deduced fromACE-MS analysis. Experimentalconditions: Fused-silica capil-lary, 80 cm (20 cm effectivelength for UV detection)650 mmID6375 mm OD; temperature257C; running electrolyte,18.561023 M ammonium ace-tate (BGE) 1 261025 M peptide1 (pH 5.1); applied voltage,125 kV; sample: mixture ofvancomycin, teicoplanin, andristocetin in BGE, concentrationof each antibiotic 561024 M;

hydrodynamic injection for 5 s; co-injection of 15.5 mM DMSO for 5 s in order to mark EOF. UV detection at 214 nm; MSdetection: sheath liquid, H2O/methanol/acetic acid (10/10/2.5 v/v/v); flow rate, 3 mL/min; nebulizing and drying gas, nitro-gen (flow rate, 210 L/h, temperature, 2007C); Vcap = 23.5 kV; Vend plate = 23.0 kV.



Figure 2. Virtual 2-D gel, corresponding to the interactions of the three antibiotics related to vanco-mycin with peptide 1, displayed by our software after ACE-MS analysis. Experimental conditions asin Fig. 1.

with peptide 1. Moreover, significant changes in theintensity of the peptide 1 MS signals can be observed atmigration times corresponding to those of the antibioticsspots and this will be discussed later on. When the con-centration of peptide 1 varied in the BGE, we observedshifts in the migration times of the spots whereas theirordinates (m/z) remain constant. This indicates, asexpected, that the stoichiometry of the interaction ofthese vancomycin related antibiotics with peptide 1 ismainly 1/1.

Thus, a single map may be sufficient to represent thewhole set of the molecular interactions. Using the dy-namic equilibrium affinity electrophoresis approach, eachmolecular system of low-affinity interaction is easily locat-ed. It is formed (i) by spots corresponding to the free andbonded ligands; (ii) by an intensity change in the receptorMS signals, which shares the same migration time as theligands; (iii) by a simple additive connection with regard tothe molecular mass of the receptor and the free ligand tovalide those of the complexes. By analogy with classical2-D gels, which permit to visualize the whole set of pro-teins present in a particular set of cells, tissue or biologi-cal fluid, we will call “virtual 2-D gel” this representation ofan ACE analysis as suggested by several authors [15–18].

3.2 Estimation of binding constants from thechange of apparent mobility

The virtual 2-D gels based upon ACE hyphenated to iontrap-MS permit to visualize the molecular interactionsoccurring in a complex solution and to identify the inter-acting partners from their mass. We want now to investi-gate whether the ion trap-MS detection can be used todetermine the binding constants from the change ofapparent mobility in a comparable way with that used inACE with the UV detection. In ACE with UV or MS detec-tion, K can be calculated from Eq. (2), assuming that thestoichiometry of the binding between ligand and receptoris 1/1. In accordance with asymptotic theory, Eq. (2) canbe rewritten as follows:

K? [R] =(mf2mi)/(mi2mc) (3)

where mf is the electrophoretic mobility of free ligand inabsence of receptor, mc is the electrophoretic mobility ofthe ligand-receptor complex, and mi the electrophoreticmobility of the ligand measured in the experimental con-ditions. Equation (3) can be transformed for a Scatchardplot:

(mi2mf)/[R] = K? (mi2mf) 1 K? (mc2mf) (4)



For a given ligand, the term (mc–mf) is in fact a constantwhatever the receptor concentration in the BGE. So,plotting (mi2mf)/[R] as a function of (mi2mf) we obtain a lin-ear relationship whose slope is the binding constant forthe studied receptor-ligand pair.

By performing successively the UV and MS detectionduring the same ACE analysis, the binding constantsrelated to peptide 1 were determined for vancomycin,ristomycin, and teicoplanin. The electrophoretic mobili-ties were calculated taking into account the EOF visual-ized in the UV or MS detection from the DMSO peak. Infact, DMSO was used as an EOF marker, as UV-absorb-ing at 214 nm, ion trap-detectable, and not interactingwith the antibiotics and their substrates. The obtainedresults are gathered in Table 1. Table 1 evidences that thecorrelation coefficients obtained for the graphical deter-minations of the binding constants by ACE-UV are sys-tematically satisfactory. In contrary, the correlation coef-ficients (R2) obtained during determination of the bindingconstants by means of ion trap-MS detection are lowerthan those corresponding to UV detection. Nevertheless,they are satisfactory enough to be able to compare thevalues obtained by the two methods. However, beforecomparing the binding constants, we have to point outthat we are working with hyphenated on-line ACE-ESI-MS whereas in the literature results are provided fromACE-UV or direct flow injection (ESI-MS). In our case, thehyphenated mass spectrometer should lead to a lowerreproducibility than direct flow injection or than ACE-UVbecause of the siphoning effect that is only partly con-trolled. Such a situation could explain the lower correla-tion coefficients observed in the literature. Assuming thispoint, we observed that whatever the studied antibiotic,the binding constants obtained graphically by using MSdetection for the peptide 1 are always higher than those

Table 1. Binding constants (M21) determined by per-forming successively UV and MS detection dur-ing the same ACE analysis

Ligand UV detection Ion trap detection

Slope Correlationcoefficient(R2)

Slope Correlationcoefficient(R2)

Vancomycin 150 000 0.994 360 000 0.94Teicoplanin 160 000 0.980 390 000 0.89Ristocetin 85 000 0.993 410 000 0.91

Results obtained graphically by plotting (mi2mf)/[R] as afunction of (mi-mf), from the ACE analyses of a mixturecontaining the three studied antibiotics (receptor: N,N-diacetyl Lys-D-Ala-D-Ala, peptide 1). Comparison be-tween UV and MS detection. Ligand concentration:561024 M; receptor concentration: 0, 5, 10, 20, 30, 40, 50,75, 100, 160, 320 mM; duplicated experiments

obtained from UV detection. This shift did not originatefrom differences in experimental conditions between ACE-UV and ACE-MS analyses. Indeed, the binding constantsevaluated from UV or MS measurements were based onthe measure of electrophoretic mobilities carried out dur-ing the same ACE analyses. The differences observed maybe ascribed to bias linked to MS detection. Moreover, thisbias has been noted from literature data. So in the case ofindependent experiments (Table 2), the binding constantof vancomycin towards peptide 1 is systematically higherfor the ESI-MS analyses (i.e., flow injection analyses)than for the ACE-UV analyses (500 000–735 000 versus240 000–375 000) [21]. In our study, taking into accountthat drying gas heats the capillary end within the nebuliz-

Table 2. Binding constants (M21) reported in literature for the three studied antibiotics with N,N-di-acetyl Lys-D-Ala-D-Ala or N-acetyl D-Ala-D-Ala as a receptor

Vancomycin [18] Teicoplanin [23] Ristocetin [23]

ESI-MS ACE-UV ACE-UV ACE-UV

N,N-Diacetyl Lys-D-Ala-D-Ala(peptide 1)

500 000;735 000a)

240 000b)

375 000c)65 000e) 100 000e)

N-Acetyl D-Ala-D-Ala (peptide 4) 19 000a) 4 500;11 000;25 000d)

140 000;37 000e)

48 000e)

a) Solutions prepared in 5 mM NH4Ac, pH 5.1; temperature, 257C [6]b) Running buffer, 20 mM sodium phosphate, pH 7.5 [24]c) Running buffer, 25 mM NH4Ac, pH 5.1 [21]d) Running buffer, 50 mM phosphate, pH 6 [27]e) Running buffer, 20 mM phosphate, pH 6.9 [28]



ing needle, this heating should alter the electrophoreticmobilities of the analyzed compounds. In fact, in the UVdetection the electrophoretic mobilities were evaluated at257C whereas the determination carried out in the iontrap-MS detection was performed at a higher tempera-ture. Such a situation probably explains the differences ofbinding constants observed between ACE–UV and ACE-ESI-MS analyses. To conclude this point, even if thebinding constants from ACE-ESI-MS and ACE-UV appearto be of the same order of magnitude, the binding con-stants from ACE-ESI-MS do not seem reliable enough tobe used.

Furthermore, it is interesting to compare the resultsobtained in ACE-UV by analyzing the mixture of the threestudied antibiotics with regard to various peptides(Table 3). Among the studied peptides, peptide 3 did notappear to form a complex with the vancomycin-relatedantibiotics. This may result from the absence of chiralrecognition between antibiotic and receptor. In fact, thelatter is a racemic compound. With respect to peptide 2,only weak interaction with teicoplanin has been evi-denced. In the literature, some binding interactions havebeen reported for these peptides leading to very weakvalues (K , 500 M21) [3, 6–9], but the binding constants ofthe antibiotics towards peptides 2 and 3 by ACE-UV werenot. This emphasizes the concentration detection limits ofthe UV and ionic trap detectors. Indeed, laser-inducedfluorescence detection [8] can provide detection limits farmore improved and should be used for very weak inter-actions studies. On the contrary, for the two other pep-tides (peptides 1 and 4) relatively strong interactions aredeveloped. Lastly, it is interesting to compare the bindingconstants obtained in ACE-UV (Tables 1 and 3) to thosereported in the literature in the case of ACE-UV analyses(Table 2). A good agreement between our results andthose reported in the literature is observed for the com-plexes formed with ristocetin involving peptide 1 orpeptide 4; the vancomycin with peptide 4 and peptide 1

although the binding constants related to the secondpeptide is slightly lower than the data reported in the lit-erature; and for the complex teicoplanin/peptide 4. Onthe contrary, concerning the complex teicoplanin/peptide1, the obtained binding constants appears significantlyhigher than the only value reported in the literature.Nevertheless, it should be noted that the pH values usedfor the two studies are relatively different (pH 5.1 in thiswork versus pH 6.9 in previous studies).

3.3 Estimation of binding constants from thechange of peak area (or spot volume)

ESI-MS has been used for studying noncovalent interac-tions [21, 29–31]. ESI is a very gentle ionization techniquethat propagates noncovalent complexes formed in solu-tion into the gas phase where they can be characterizedby MS analysis. The unique contribution of the MSdetectors is that it allows the electropherogram of eachmolecular species to be extracted in such a way that thisgives access to the individualization of ligands, receptors,and complexes (Fig. 3), instead of being joined under thesame peak with the UV detection (Fig. 4). Thus, for theACE-ESI-MS methodology, the estimate of the bindingconstants from the change of peak area could appear asa reachable objective.

The binding constant corresponding to the 1/1 molecularassociation between a ligand and its receptor can becalculated from Eq. (1) as far as the concentrations ofligand, receptor, and complex at equilibrium can bedetermined. However, the concentration of molecularspecies is not directly accessible since ion trap measuresonly ionic intensities. Consequently, one must focus onthe real significance of the ionic signals detected in massspectrometers in connection with the binding constantsexisting in biological conditions [22, 32]. The primaryquestion that must be asked is the following: as the threemolecular species (ligand, receptor, and complex) are

Table 3. Binding constants (M21) determined graphically from ACE-UV analyses of a mixture con-taining the three studied antibiotics

Vancomycin Teicoplanin Ristocetin

N,N-Diacetyl Lys-D-Ala-D-Lac (peptide 2)

NC 500 NC

N-Acetyl-Ala-Ala-Ala (peptide 3) NC NC NCN-Acetyl D-Ala-D-Ala (peptide 4) 11 000

(R2 = 0.997)45 000(R2 = 0.98)

70 000(R2 = 0.98)

NC, no complex detectedLigand concentration: 561024 M; receptor concentration: 0, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100,400 mM; triplicated experiments



Figure 3. Extracted ACE-UV-MS electropherograms of vancomycin interacting with peptide 1.Experimental conditions as in Fig. 1 except running buffer: 18.561024 M ammonium acetate (BGE) 1261025 M peptide 1 (pH 5.1); sample, 561024 M vancomycin in BGE.

Figure 4. Electropherogram recorded in UV during theACE-UV-ESI-MS analysis reported in Fig. 3. Experimentalconditions as in Fig. 3.

present at the same time in ion trap, can we link the rela-tive abundance of these three ion signals to their molec-ular concentration? Furthermore, environmental condi-tions inside the ion trap (gas phase transition, vacuum,temperature, high voltages, etc.) are far removed fromthose existing in the buffers that fill the capillaries during

the electrophoresis runs. The next question is thereforewhether the association constants that are derived underthese conditions reflect the bindings occurring in solution.Indeed, studying noncovalent interactions in aqueousmedia by ACE-UV and in media short of water with MSdetection should be quite complementary concerning thenature of the affinity [14, 33]. The straightforward obser-vation of Figs. 1 and 2 shatters these speculations: a sig-nificant increase of the receptor ionization intensity issystematically seen at the migration time correspondingto each antibiotic system whereas the initial concentra-tion of the receptor should be reduced at equilibrium. Thisgives the answer to the first question and leads to con-clude that the virtual 2-D gels based upon ACE hyphe-nated to ion trap-MS cannot allow one to estimate thebinding constants of low-affinity molecular systems fromthe change of spot volumes.

4 Concluding remarks

Taking advantage of the robustness, efficiency, and easeof use of ACE for the analysis of biomolecular non-covalent interactions and of the selectivity of ion trap-MS



used as detector, we carried out the coupling on-line ofthese two separation technologies. We also developed asoftware application, which was designed to create andanalyze the virtual 2-D gels. The virtual 2-D gels are ima-ges built from data issued from signals transmitted by anion trap-MS hyphenated to ACE. They are the raw inputdata to our software programs from which the interactingmolecular species may be detected, quantified, and theiraffinity measured. The validity of this new analytical toolfor probing biomolecular interactions has been demon-strated on mixtures of antibiotics of the vancomycingroup and several dipeptide substrates. Using the dy-namic equilibrium affinity electrophoresis approach forACE separation, the molecular components that interactwith a low affinity are easily located on the virtual 2-D gelsand the stoichiometry of interaction can be assessed. Wehave shown that the binding constants of the whole set ofinteracting molecules can be estimated from the changeof apparent mobilities. For that purpose, UV detectionshould be used together with the MS detection as resultsobtained from MS are systematically overvalued. Ourresults concerning the determination of affinity from thechange of peak area (or spot volume) demonstrated thatour ACE-ESI-MS approach is not at all adapted for low-to-intermediate affinity interactions. We still have toestablish whether it is adapted for the determination ofstrong affinity bindings.

It appears that the novel analysis approach we havedeveloped could be a very efficient research tool for thecharacterization of the whole set of molecular interactionsoccurring in a complex mixture of biological compounds.A closely related technique to it is the biomolecular inter-action analysis-mass spectrometry approach (BIA-MS)[34]. Similarly to ACE-MS, it combines a bioaffinity analy-sis tool, using the surface plasmon resonance phenom-enon for characterizing macromolecular interaction [35],with identification of interacting proteins by MS. However,the coupling is on-line for our ACE-MS approach,whereas it is off-line for BIA-MS consisting of successivesteps, binding of an analyte in solution to its ligandimmobilized on the sensor surface, optical detection,microelution procedure, tryptic digestion, and MALDI-TOF-MS (or MS/MS) analysis and database searches forprotein identification. The potential of these two approa-ches is similar but not exactly identical. BIA-MS permitsto determine the kinetic constants of the interactionswhereas ACE-MS only gives access to the binding con-stants. The former identifies interacting proteins while thelatter, in the present state of its development, only char-acterizes the interacting molecules by the mass. The for-mer can only deal with one analyte at the same time whenour methodology permits to take into account the wholeset of interactions in a complex mixture.

For this last reason, the analysis of virtual 2-D gels basedupon affinity CE hyphenated to ion trap-MS has thepotential of becoming one of the significant tools of func-tional proteomics research. Further developments cer-tainly are needed to really fulfil these promises. Theadsorption of polypeptides to the inner surface of capil-laries still remains a major problem for binding studiesusing ACE. The noncompatible nature of usual CE andMS buffers also is an impediment for an optimal couplingof these techniques [36]. Exploring all the differentapproaches offered by CE to quantify the molecularinteractions, especially the frontal analysis methodology[3, 37], in view of combining them with MS detection, hasstill to be done. Our dedicated software is still in itsinfancy: we are currently specifying acceptance criteriafor the automatic spot and line detection and developingan integrated system would without doubt be welcome.

Received September 22, 2004

5 References

[1] Chu, Y.-H., Avila, L. Z., Biebuyck, F. A., Whitesides, G. M., J.Med. Chem. 1992, 35, 2915–2917.

[2] Heegaard, N. H. H., Robey, R. T., Anal. Chem. 1992, 64,2479–2482.

[3] Guijt-van Duijn, R. M., Frank, J., van Dedem, G. W. K., Bal-tussen, E., Electrophoresis 2000, 21, 3905–3918.

[4] Heegaard, N. H. H., Kennedy, R. T., Electrophoresis 1999,20, 3122–3133.

[5] Heegaard, N. H. H., Nissen, M. H., Chen, D. D., Electropho-resis 2002, 23, 815–822.

[6] Van Der Kerk-Van Hoof, A., Heck, A. J. R., J. Antimicrob.Chemother. 1999, 44, 593–599.

[7] Van Der Kerk-Van Hoof, A., Heck, A. J. R., J. Mass. Spec-trom. 1999, 34, 813–819.

[8] Propieniek, P. H., Pratt, R. F., Anal. Biochem. 1987, 165,108–113.

[9] Allen, N. E., Le Tourneau, D. L., Hobbs Jr., J. N., J. Anti-biot.1997, 50, 677–684.

[10] Peng, X., Bowser, M. T., Britz-McKibbin, P., Bebault, G. M.,Morris, J. R., Chen, D. Y., Electrophoresis 1997, 18, 706–716.

[11] Dunayevskiy, Y. M., Lyubarskaya, Y. V., Chu, Y. H., Vouros, P.,Karger, B. L., J. Med. Chem. 1998, 41, 1201–1204.

[12] Lynen, F., Zhao, Y., Becu, C., Borremans, F., Sandra, P.,Electrophoresis 1999, 20, 2462–2474.

[13] Morin, C. J., Geulin, L., Desbène, A., Desbène, P. L., J.Chromatogr. A 2004, 1032, 327–334.

[14] Guzman, N. A., Stubbs, R. J., Electrophoresis 2001, 22,3602–3628.

[15] Breme, U., Breton, J., Visco, C., Orsini, G., Righetti, P. G.,Electrophoresis 1995, 16, 1381–1384.

[16] Loo, J. A., Brown, J., Critchley, G., Mitchell, C., Andrews, P.C., Ogorzalek Loo, R. R., Electrophoresis 1999, 20, 743–748.

[17] Walker, A. K., Rymar, G., Andrews, P. C., Electrophoresis2001, 22, 933–945.



[18] Lubman, D. M., Kachman, M. T., Wang, H., Gong, S., Yan, F.,Hamler, R. L., O’Neil, K. A., Zhu, K., Buchanan, N. S., Barder,T. J., J. Chromatogr. B 2002, 782, 183–196.

[19] Colton, I. J., Carberck, J. D., Rao, J., Whitesides, G. M.,Electrophoresis 1998, 19, 367–382.

[20] Ge, M., Chen, Z., Oninshi, H. R., Kohler, J., Silver, L. L.,Kerns, R., Fukuzawa, S., Thompson, C., Kahne, D., Science1999, 284, 507–511.

[21] Jørgensen, T. J. D., Roepstorff, P., Heck, A. J. R., Anal.Chem. 1998, 70, 4427–4432.

[22] Vollmerhaus, P. J., Tempels, F. W. A., Kettenes-van denBosch, J. J., Heck, A. J. R., Electrophoresis 2002, 23, 868–879.

[23] Bonnici, P. J., Damen, M., Waterval, J. C. M., Heck, A. J. R.,Anal. Biochem. 2001, 290, 292–301.

[24] Chu, Y.-H., Whitesides, G. M., J. Org. Chem. 1992, 57,3524–3525.

[25] Sharman, G. J., Try, A. C., Dancer, R. J., Cho, Y. R., Star-oske, T., Bardsley, B., Maguire, A. J., Cooper, M. A., O’Brien,D. P., Williams, D. H., J. Am. Chem. Soc. 1997, 119, 12041–12047.

[26] Geulin, L., Thesis, University of Rouen, 2002.

[27] Busch, M. H. A., Boellens, H. F. M., Kraak, J. C., Poppe, M.,J. Chromatogr. A 1997, 775, 313–326.

[28] Azad, M., Silveirio, C., Zhang, Y., Villarial, V., Gomez, F. A., J.Chromatogr. A 2004, 1027, 193–202.

[29] Lim, H. K., J. Mass Spectrom. 1995, 30, 708–714.

[30] Loo, J. A., Mass Spectrom. Rev. 1997,16, 1–23.

[31] Loo, J. A., Sannes-Lowery, K. A., Mass Spectrom. Biol.Mater. 1998, 345–367.

[32] Kempen, E. C., Brodbelt, J. S., Anal. Chem. 2000, 72, 5411–5416.

[33] Rabilloud, T., Electrophoresis 1996, 17, 813–829.

[34] Lopez, F., Pichereaux, C., Burlet-Schiltz, O., Pradayrol, L.,Monsarrat, B., Esteve, J. P., Proteomics 2003, 3, 402–412.

[35] Szabo, A., Stolz, L., Granzow, R., Curr. Opin. Chem. Biol.1995, 5, 699–705.

[36] Heegaard, N. H. H., Electrophoresis 2003, 24, 3879–3891.

[37] Østergaard, J., Heegaard, N. H. H., Electrophoresis 2003,24, 2903–2913.


Documents

Analysis of virtual two-dimensional gels based upon affinity capillary electrophoresis hyphenated to ion trap-mass spectrometry