ARTICLE pubs.acs.org/ac
Quantifying ProteinProtein Interactions Within Noncovalent Complexes Using Electrospray Ionization Mass Spectrometry Elisabetta Boeri Erba, Konstantin Barylyuk, Yang Yang, and Renato Zenobi* Department of Chemistry and Applied Biosciences, ETH Zurich, CH-8093 Zurich, Switzerland
bS Supporting Information ABSTRACT: Several electrospray-mass spectrometry (ESI-MS)based methods are available for determining the constant of association (Ka) between a protein and a small ligand, but current MS-based strategies are not fully adequate for measuring Ka of proteinprotein interactions accurately. We expanded the application of ESI-MS-based titration to determine the strength of noncovalent interactions between proteins, forming a complex. Taking into account relative response factors (probability of being ionized, transmitted, and detected), we determined Ka values of an equilibrium between dimers and tetramers at three different pH values (6.8, 3.4, and 8.4). We investigated the association of the lectin concanavalin A, whose dimertetramer ratio in the gas phase is affected by solution concentration and by pH. To calculate the constants of association in solution, we also utilized isothermal titration calorimetry (ITC) for a comparison with MS-based titration. At pH 6.8 and pH 8.4, the Ka values measured by MS and by ITC were in agreement. ITC results allowed us to restrain the response factor to a value close to 4. At pH 3.4, we were able to measure the Ka only by MS, but not by ITC because of limited sensitivity of calorimetry. Our investigation illustrates the great potential MS for calculating the binding strength of proteinprotein interactions within noncovalent complexes. The main advantages of MS over ITC are its sensitivity (i.e., the required amount of sample is >100 times less than the one necessary for ITC), and the possibility to obtain precise information on composition of protein complexes, their stoichiometry, their subunit interactions, and their assembly pathway. Compared to previous investigations, our study shows the strong influence of response factors on determining accurate proteinprotein association constants by MS.
’ INTRODUCTION Regulated interactions between proteins to form noncovalent complexes are essential in many cellular processes, such as signal transduction pathways, replication of DNA, and activity of enzymes.1 These interactions have many diverse roles and their quantification is of pivotal importance to better understand biological events and drugtarget interactions. Traditionally, isothermal titration calorimetry (ITC) has been used to quantitatively measure the binding affinity of proteins to their ligands.2 The self-association of a protein, which leads to the formation of homocomplexes, has been also analyzed by ITC in cases where monomerdimer equilibria have been investigated.35 Recently, electrospray mass spectrometry (ESI-MS) has emerged as an alternative to ITC and other biophysical techniques, because of its sensitivity and speed of analysis. The MSbased approaches to quantify noncovalent interactions can be divided in two categories, titration methods, and competition methods.6 In the former case, a biomolecule of interest (B) is kept at a fixed concentration and the amount of an interacting partner (ligand, L) is gradually raised, forming an increasing amount of a B 3 L complex.710 In the latter case, B is incubated with L and a second ligand, a competitor (C), is added.1113 In many MS-based approaches, the ratio of signal intensities of the different ions in the spectra is considered equal to the ratio of r 2011 American Chemical Society
concentration of analytes in solution. In other words, it is assumed that B 3 L and B have the same probability of being ionized, transmitted, and detected. However, in general, B and B 3 L may not be analyzed with the same efficiency due to their different composition and masses. In more rigorous MS-based approaches, the distinct ionization, transmission and detection probability of B and B 3 L are taken into consideration and quantified. This means that a correlation between their solution concentration and their signal intensities is calculated. Gabelica et al. developed an elegant method to determine the equilibrium association constant (Ka) of a complexation reaction between L and B in equimolar concentration to form B 3 L.14 The fitting procedure allowed the simultaneous determination of the Ka and a factor R. The latter is the ratio between the response factors of B 3 L and B (RB 3 L and RB). A response factor is the product of efficiencies of all the processes affecting the signal intensities of the ions. It takes into account the distinct ESI responses (ionization probability) of B 3 L and B, the differences in instrumental transmission and detection efficiencies, and the possible in-source dissociation of the complex B 3 L. The group of Received: June 21, 2011 Accepted: November 2, 2011 Published: November 02, 2011 9251
dx.doi.org/10.1021/ac201576e | Anal. Chem. 2011, 83, 9251–9259
Analytical Chemistry
ARTICLE
Figure 1. Relative abundance of dimers and tetramers versus solution concentration of concanavalin A (a) Spectrum of intact concanavalin A at an analytical concentration D0 = 20.2 μM at pH = 6.8; (b) spectrum of intact concanavalin A at D0 = 2.1 μM and pH 6.8. As the concentration of concanavalin A decreased, the tetramer becomes less abundant. Two spheres indicate the dimeric ions and four spheres indicate tetrameric ions. (c and d) Plots of the ratio of areas under the peaks representing the dimeric (ID) and the tetrameric (IT) populations (ID/IT) versus analytical concentration D0 at pH 6.8. The factor R, which is ratio between the response factors of the corresponding analytes (RD and RT, see equation 3a), was restricted to values between 0.5 and 6. The data fitting allowed the determination of the constant of association (Ka) of dimertetramer interaction. The dotted lines in the plot indicate a 95% confidence interval.
Gross presented a method for determining the affinity constants and response factors for weakly bound noncovalent complexes that could dissociate during the ESI process.7 Compared to Gabelica’s approach, Gross’ method does not require a precise determination of concentration of interacting molecules. Gross’ study focused on the dimerization of the pentadecapeptides Gramicidin D. This is one of the few cases where researchers investigated the strength of the interactions between polypeptides by ESI-MS. Ayed et al. also determined the Ka of protein protein interactions, studying the equilibrium between the dimeric citrate synthase and its hexameric form.15 They measured Ka values for proteinprotein interactions and constants of dissociation (Kd) for proteinligand, taking into account the different sensitivity of their instrument for the dimers and hexamers. Very recently, Liu and Konermann studied protein protein binding affinities by ESI-MS using β-lactoglobulin and hemoglobin.16 They showed that there was no conformational change during proteinprotein binding for the biomolecules they analyzed. There was also no disruption of noncovalent interactions during ion transport and no formation of nonspecific interactions. On the basis of this information, they concluded that response factors of monomers and dimers, which they investigated, were very similar. Here we report the use of nano-ESI-MS to study the association of a noncovalent homo-oligomeric complex, concanavalin A, whose dimeric form is in equilibrium with its tetramer. Concanavalin A is a lectin from jack-bean (Canavalia ensiformis) that specifically binds saccharides containing α-D-mannose and α-Dglucose residues.17 For this property, it is considered a model
system for the study of proteinoligosaccharide interactions. Because of its specific binding properties, it has been utilized in a biosensor where concanavalin A binding to the E. coli surface mannose enhanced the adhesion of the bacteria to a mannosemodified magnetoelastic sensor.18 It is also used to purify glycosylated macromolecules in lectin affinity chromatography.19 Concanavalin A exhibits a specific cellular reactivity because of its binding to surfaces of cells: it induces agglutination (i.e., clumping) of cells. It stimulates cell division and because of its mitogenic activity it is considered as an archetype of lectins having anticancer properties.20 Each mature subunit of concanavalin A is composed of 237 amino acids, whose theoretical mass is 25598 Da, and contains two metal binding sites, which are necessary for binding saccharides.21 In solution, concanavalin A is able to self-associate and is present as dimers or tetramers depending on pH, concentration and temperature.17 In the early 1980s, Teller and co-workers studied the dimertetramer equilibrium of concanavalin A using analytical ultracentrifugation, indicating that this lectin was in its dimeric form at pH values below 5.5. They also suggested that concanavalin A showed an increasing ability of self-assembly up to pH 7.5.22 Although this protein has been extensively used and studied, no recent investigation of the concanavalin A oligomeric states in solution has been carried out and the dimer-tetramer equilibrium has not been rigorously characterized. We analyzed the equilibrium between the distinct oligomeric states of concanavalin A, and we determined the Ka values of the dimer-tetramer equilibrium using ESI-MS-based titration at 9252
dx.doi.org/10.1021/ac201576e |Anal. Chem. 2011, 83, 9251–9259
Analytical Chemistry three different pH values (6.8, 3.4, and 8.4). For comparison, we also utilized ITC to calculate the Kd in solution.
’ EXPERIMENTAL SECTION Materials. Ammonium acetate (NH4Ac) and concanavalin A were purchased from Sigma-Aldrich (Buchs, Switzerland). Ammonia was from Fluka (Buchs, Switzerland). Acetic acid was purchased from Carlo Erba (Rodano, Italy). Nanoflow platinumcoated borosilicate electrospray capillaries were bought from Proxeon (Odense, Denmark). Sample Preparation for Mass Spectrometry. As buffer we used 100 mM NH4Ac at pH 3.4, pH 6.8, and pH 8.4. The pH was adjusted to 3.4 and to 8.4 by adding acetic acid or ammonia, respectively. Solutions of concanavalin A in NH4Ac were prepared at 200 μM concentration, then desalted and buffer exchanged using ultrafiltration (Vivaspin 500, Sartorius Stedim Biotech SA, Tagelswangen, Switzerland) with a 10 kDa cutoff. After buffer exchange, the concentration of concanavalin A was determined by measuring the absorbance at 280 nm using an UV/vis spectrophotometer (LAMBDA 20, PerkinElmer, Schwerzenbach, Switzerland) and by using Beer’s law. The extinction coefficient of concanavalin A (32430 M1 cm1) was calculated online using the ExPASy Proteomics tools. To exclude the possible adsorption of concanavalin A to the inner surface of vials, we weighted known amounts of concanavalin A, dissolved the different aliquots of protein in 100 mM NH4Ac (without any buffer exchange) and measured their absorbance at 280 nm; no loss of material because of adsorption to surfaces was detected. Prior to mass spectrometric analysis, the concanavalin A solutions were diluted in NH4Ac at the desired pH to the concentrations indicated in Figures 1c, 3a, and 3b. To generate denaturated concanavalin A to further control the purity of the protein complex sample (data non shown), we utilized Ziptip columns containing C4-resin and C18-resin (Millipore, Molsheim, France) as previously described.23 Mass Spectrometry. Instrumental Settings. Protein complex ions were generated using a nanoflow electrospray (nano-ESI) source. Mass spectrometry analyses were carried out on a quadrupole time-of-flight mass spectrometer (Q-TOF Ultima, Waters Corporation, Manchester, U.K.). The instrument was modified for the detection of high masses.24,25 The following instrumental parameters were used: capillary voltage up to 2 kV, cone potential = 40 V, RF lens-1 potential = 40 V, RF lens-2 potential = 1 V, aperture-1 potential = 0 V, collision energy = 20 V, and microchannel plate (MCP) = 200 V. The ion transmission was optimized for a m/z range between 2000th and 8000th using the following setting of the quadrupole (named as “MS profile”): mass 3000, 0% dwell scan time, 50% ramp scan time; mass 4000, 0% dwell scan time, 50% ramp scan time. All mass spectra were calibrated externally using a solution of cesium iodide (2 mg/mL in 50% isopropanol) and were processed with the Masslynx 4.0 software (Waters Corporation, Manchester, U.K.). Titration Experiments and Data Fitting. To evaluate the MSbased titration data below, we need the total concentration of the concanavalin A dimer, D0, which we call “analytical concentration”.26 This was measured using UVvis spectrophotometry. The “equilibrium concentration” is the real concentration of dimers and tetramers when they are at the equilibrium in solution.
ARTICLE
To carry out the MS-based titration experiments, the analytical concentration of concanavalin A, loaded into the nano-ESI needles, was varied over a fairly wide range (D0 between 0.3 and 22.1 μM), resulting in significant changes of the relative abundance of tetramer and dimer. For each concentration, at least three spectra were acquired and around thirty different concentrations were analyzed. At each concentration, the acquired mass spectra were smoothed using the Masslynx software. Then, the peak areas representing the dimeric (ID) and the tetrameric (IT) populations were calculated. The ratio of ID to IT (ID/IT) was plotted versus the analytical concentration of the dimer (D0) (Figure 1c, 3a, and 3b). By using the software MATLAB_R2010a (Mathworks, Natick, MA, U.S.A.), eq 4 was utilized to fit the association constant Ka and the factor R, which is ratio between the response factors of the corresponding analytes (eq 3a). RD and RT account for the differences in electrospray ionization efficiency of the dimers and tetramers, in their instrumental transmission, in their detection, and their possible in-source dissociation. For the derivation of eq 4, see Appendix I. 2D S T Ka ¼
R¼
ð1Þ
1 ½T ¼ K d ½D2
ð2Þ
RT RD
ð3aÞ
I D R D ½D 1 ½D ¼ ¼ I T R T ½T R ½T pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi I D 1 þ 1 þ 8K a D0 ¼ 2RKa D0 IT
ð3bÞ
ð4Þ
To calculate the solvent excluded surface (SES) of dimeric and tetrameric concanavalin A, we used the structure 1CON21 in the PDB database and the software StrucToolsPDB coordinate calculations http://helixweb.nih.gov/structbio/basic.html. Isothermal Titration Calorimetry (ITC). Sample Preparation for ITC. Concanavalin A was dissolved in 100 mM NH4Ac buffer at pH 6.8 or pH 8.4 at concentrations between 150 and 200 μM. The exact concentration of concanavalin A was determined by measuring the absorbance at 280 nm. ITC Experiments. We utilized a commercial isothermal titration calorimeter (VP-ITC system, MicroCal, GE Healthcare Europe GmbH, Glattbrugg, Switzerland) to perform “dilution experiments”.2 In these experiments, a solution of concanavalin A (in 100 mM NH4Ac at pH 6.8 or pH 8.4) was loaded into the syringe of the ITC apparatus and a buffer solution (100 mM NH4Ac, at pH 6.8 or pH 8.4) was in the calorimetric cell. During each ITC experiment, between 7 and 10 injections of concanavalin A into the calorimetric cell were carried out. The duration of each injection was 5070 s and the time between the injections was 400500 s. The volume of each injection was between 25 and 35 μL, the stirring speed was 480 rpm. The ITC experiments at each pH were repeated four times to verify the reproducibility of the measurements. The stepwise addition of the concanavalin A solution into the calorimetric cell determined the dilution of the tetramer and triggered its dissociation into dimers. The propensity of tetramers 9253
dx.doi.org/10.1021/ac201576e |Anal. Chem. 2011, 83, 9251–9259
Analytical Chemistry
ARTICLE
Table 1. Determination of Ka by MS-Based Titration and ITC at pH 3.4, pH 6.8, and pH 8.4a by ESI-MS
pH = 3.4
pH = 6.8
1
R
pH = 8.4
1
Ka (M1)
Ka (M )
Ka (M ) 0.5
4.0 104 ( 0.5 104
51.0 104 ( 0.4 104
1
1.8 10 ( 0.2 10
18.0 10 ( 0.1 10
140 104 ( 10 104
2
0.85 10 ( 0.09 10
7.3 10 ( 0.4 10
53.0 104 ( 3.0 104
4
0.41 10 ( 0.04 10
3.2 10 ( 0.1 10
23.0 104 ( 0.1 104
6 by ITC
4
4
4
4
4
4
4
4
0.27 10 ( 0.05 10 not applicable 4
4
4
4
4
2.1 10 ( 0.1 10 4.1 104 ( 1.8 104
4
4
4
390.0 104 ( 20 104
14.5 104 ( 1.0 104 21.0 104 ( 12.0 104
Using MS-based titration method and the fitting procedure previously described (eq 4), we measured Ka values of the dimer-tetramer equilibrium of concanavalin A in solution at pH 3.4, pH 6.8, and pH 8.4. Using dilution method and ITC, we determined the Kd of tetramers dissociating into dimers in solution at pH 6.8 and pH 8.4. The low abundance of tetramers at pH= 3.4 did not allow the determination of Kd at this pH value. The shown Ka values are the inverse of Kd experimentally determined by ITC. The Ka values are expressed in M1, using the same exponentiation (104). When the factor R is equal to 4 (indicated in bold), there is an excellent agreement between Ka_MS and Ka_ITC. a
to dissociate is governed by its dissociation constant Kd. We used the Origin 7.0 software, provided by MicroCal, to determine Kd of the dimertetramer equilibrium. The heat qi associated with each injection of the tetrameric complex into the reference cell is proportional to the increment in the equilibrium concentration of dimer [D]i in the calorimetric cell after the injection i, because of the dissociation of the tetramer:2 v qi ¼ V ΔH t ½Di ½Di1 F 0 ½D0 ð5Þ V where V is the calorimetric cell volume, ΔHt is the enthalpy of dissociation (per dimer), [D]i is the concentration of dimer after injection i, F0 is the fraction of dimer already present in the concentrated solution placed in the syringe, [D]0 is the total analytical concentration of dimer in the syringe, and v is the injection volume. The heat qi released or absorbed during injection i, proportional to the change in concentration of dimer after the injection i ([D]i [D]i1) [see eq 5]. The term F0[D]0(v/V) is a correction factor that takes into accounts the increment of dimer concentration in the cell due to the amount of dimer already present in the syringe, which does not contribute to the heat qi.2 By iterative fitting of eq 5, [D]i and [D]i1 are obtained. By expressing [D]i as a function of the total analytical concentration of dimer in the cell after the injection i, [D]T,i provides Kd [see eq 6]. Kd ¼
Kd ¼
½DT, i ½Di 1 ½Di 2 ¼ ½DT, i ¼ ½Di þ 2½Ti ½Ti ¼ ½Ti 2 Ka
½Di 2 2½Di 2 ¼ ½DT, i ½Di ½DT, i ½Di 2
2½Di 2 þ K d ½Di K d ½DT, i ¼ 0 2 ½Di 2 þ ½Di ½DT, i ¼ 0 Kd 0sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 8½DT, i Kd@ ½Di ¼ 1A 1 þ 4 Kd
ð6Þ
The Kd values of the dimertetramer equilibrium, obtained by ITC, were converted to Ka values (Ka = 1/Kd, Table 1). The error (ΔKa) of the Ka values was calculated from the ΔKd (calculated by Origin 7.0 software) as follows: ΔK a ¼
1 ΔK d K d2
’ RESULTS AND DISCUSSION Oligomeric states of gas-phase concanavalin A is influenced by the solution concentration. In the available literature
there is an early study where concanavalin A was investigated at different pH values using analytical ultracentrifugation.22 Using nano-ESI-MS we wanted to study the oligomeric states of concanavalin A in native condition to update the information on this lectin. The mass spectra indicate that at a dimer analytical concentration of D0= 22.2 μM, at pH= 6.8 in 100 mM NH4Ac, concanavalin A is detected as tetramer (T) and as dimer (D) whose experimental masses are 102392 ( 7 Da and 51196 ( 5 Da, respectively (Figure 1a). The charge states of concanavalin A tetrameric ions range from 18+ to 21+, while those of dimeric ions are between 13+ and 16+ (Figure 1a and 1b). At D0 = 20.2 μM the intensity of the 20+ tetrameric ion is 82% of the 14+ dimeric ion (Figure 1a). At D0 = 2.1 μM at pH = 6.8 in 100 mM NH4Ac, concanavalin A is approximately ten times less concentrated than at 20.2 μM, and the intensity of the tetrameric ion 20+ is half of the intensity of the 14+ dimeric ion (Figure 1b). This indicates that the abundance of the two oligomeric species in the gas phase reflects the dimer-tetramer equilibrium of concanavalin A in solution.27,28 To correlate the relative intensities of dimers and tetramers in the mass spectra and the relative abundance of the different species in solution, response factors of dimers and tetramers should be taken into account during the analysis of the MS data (see below). To quantify the strength of the interactions between two dimers of concanavalin A forming a tetramer at pH 6.8, we used a MS-based titration. The concentration of concanavalin A was gradually increased and the equilibrium between the two distinct oligomeric states was found to shift toward the tetramer (Figure 1a and 1b). For each concentration we calculated the areas under the peaks representing the dimeric (ID) and the tetrameric (IT) populations. We plotted the ratio of ID to IT 9254
dx.doi.org/10.1021/ac201576e |Anal. Chem. 2011, 83, 9251–9259
Analytical Chemistry
ARTICLE
Figure 2. Influence of solution pH on the dimers and tetramers of concanavalin A. (a) Spectrum of intact concanavalin A at D0 = 18.3 μM and pH 3.4; (b) spectrum at D0 = 2.1 μM and pH 3.4; (c) spectrum of concanavalin A at D0 = 16.9 μM and pH 8.4; (d) spectrum of concanavalin A at D0 = 2.0 μM and pH 8.4. Two spheres indicate the dimeric ions and four spheres indicate tetrameric ions.
(ID/IT) versus D0 (Figure 1c); as D0 increases, the abundance of IT grows and ID/IT decreases. We performed a double-parametric nonlinear fitting of the ESI-MS data using eq 4 (see Experimental Section and Appendix I) (Figure 1c). We determined the constant of association (Ka) of dimertetramer interaction at pH 6.8, taking into consideration the factor R, (eq 3a), as previously described.14 Gabelica et al. emphasized that the fitting of a titration curve requires a very high quality data set.14 In other words, the data points should align along the fitting curve in a nearly ideal way to allow the precise determination of both factor R and Ka (see Figure 1 of Gabelica et al.14). However, the experimental data, acquired during our nano-ESI analyses of concanavalin A, showed ID/IT scattered (for a variety of reasons, e.g., instability of the nano-ESI spray and peak broadening due to solvent adducts). Therefore, when we fitted the data using unconstrained values for factor R and Ka, we obtained very broad confidence intervals for the calculated parameters. To overcome this problem and carry out more accurate data analysis, we decided to restrict the range of the factor R. To limit R, we first consulted the literature where the factor R for biomolecules was reported to range between 0.5 and 6.7.1416,27 Second, we considered the number of charge and solvent excluded surface (SES) of the dimer and tetramers (see also the section titled “Comparison of Ka Values Determined by MS and by ITC”). Taking into account this information on dimeric and tetrameric concanavalin A and the values reported in the literature, we hypothesized that the factor R would be larger than 1, but below 6.7. By nonlinear fitting and computing a 95% confidence interval, we obtained Ka values at pH 6.8, which range between (2.0 ( 0.1) 104 M1 and (51 ( 4) 104 M1, depending on the choice of the factor R (Table 1 and Figure 1c). These data are consistent with the Ka obtained by analytical ultracentrifugation (56 104 M1).22 Analysis of Oligomeric States of Concanavalin A as a Function of Solution pH. As shown above, we monitored the
dimertetramer equilibrium of concanavalin A at pH 6.8 as D0 was increased (Figure 1a and 1b). Afterward, we wanted to verify whether the oligomeric state of concanavalin A was not only influenced by the solution concentration, but also by the pH of the NH4Ac buffer in which concanavalin A was dissolved. Therefore, we analyzed concanavalin A at pH 3.4 and pH 8.4 (Figure 2). At both pH values we were able to detect both dimers and tetramers, but the relative abundance of dimer and tetramer was very different. At pH 3.4, the dimeric state was strongly favored (Figure 2a and b). At D0 = 18.3 μM the intensity of the 20+ tetrameric ions was only 18% of the intensity of the 14+ dimeric ion (Figure 2a). At D0 = 2.1 μM the tetramer was hardly detectable (Figure 2b). The presence of monomer at pH 3.4 was as insignificant as at pH 6.8, indicating that the low pH had no effect on the dissociation of the dimer. At pH 8.4, D0 = 16.9 μM the tetramer was more abundant than the dimer: the intensity of the 20+ tetrameric ions was 120% of the height of the 14+ dimeric ions (Figure 2c and 2d). This change in the abundance of the oligomeric species according to the pH can be explained by considering the interactions involved in the tetramer formation.21 Between two dimers of concanavalin A forming a tetramer, the crystallographic structure indicated that there are 18 proteinprotein hydrogen bonds, 20 hydrogen bonds involving a bridging water molecule, and 114 van der Waals contacts (see Table 6 in Naismith et al.21). To understand why, according to our results, the tetramer is more prevalent at pH 8.4 than at pH 3.4, we made the following simple analysis. We assumed that the nature of favorable intermolecular interactions is positively correlated with the intermolecular binding energy. We further assumed that the pKa values of the individual amino acids are unchanged by the protein environment. By making these simplifications, we qualitatively evaluated the nature (i.e., hydrogen bonds, ionic bonds and van der Walls contacts) of the binding between amino acids involved in the formation of the tetramer. A larger number of strong 9255
dx.doi.org/10.1021/ac201576e |Anal. Chem. 2011, 83, 9251–9259
Analytical Chemistry
ARTICLE
Figure 4. Dissociation of tetramers into dimers observed by ITC. The Kd values of dimer-tetramer equilibrium (a) at pH 6.8 and (b) pH 8.4 were determined using the eqs 5 and 6 (see Experimental Section) to fit the ITC data. The dissociation of concanavalin A tetramers into dimers is exothermic.22
Figure 3. Determination of Ka at pH 3.4 and pH 8.4. (a) Plot of the ratio of areas under the peaks of dimers (ID) and tetramers (IT) (ID/IT) versus analytical concentration D0 at pH 3.4 and (b) plot of ID/IT versus D0 at pH 8.4.
proteinprotein interactions exist at pH 8.4 than at pH 3.4. At basic pH, the acidic residues (Asp58, Glu192, see Table 6 in Naismith et al.21) are deprotonated (1 charge), the basic amino acids (Arg60, Lys114, Lys116) are protonated (+1 charge), and polar residues (Ser62, Ser66, Ser72, Ser76, Ser108, Ser109, Ser117, Thr74, Thr120, Thr194, Thr196; Tyr67, Asn69, Asn118) and His residues (His51, His121) present no charges. In contrast, at pH 3.4 the acidic residues become protonated (i.e., no charge), the basic amino acids and His are protonated (+1 charge). When we examined the bonds, involving the amino acids mentioned above, we observed a change in the nature of the interactions according to pH (see Supporting Information Tables 1 and 2). For instance, crystallographic data indicated a hydrogen bond between Asp58 and Arg60. At pH 8.4 the deprotonated Asp58 is able to form a salt bridge with the protonated Arg60, making the interaction Asp58Arg60 at basic pH much stronger than at pH 3.4. At this pH, the noncharged Asp58 could bind to the protonated Arg60 via a charge-assisted hydrogen bond. Another interesting example is the interaction His51-Lys116, which is indicated as hydrogen bond through a single bridging water by the crystallographers.21 At pH 8.4, His51 presents no charges and Lys116 is positively charged, forming a charge-assisted hydrogen bond. On the contrary, at pH 3.4, both these basic amino acids are positively charged inducing a repulsion. Overall, we suggest a simple explanation for the relative abundance of the dimers and tetramers at different two pH values by considering protonation and deprotonation of the amino acids, whose interactions are involved in the tetramer formation.
After verifying that the oligomeric state of concanavalin A was influenced by the solution pH, we quantified the strength of the interactions forming tetramers of concanavalin A at pH 3.4 and pH 8.4. Using MS-based titration and the fitting procedure previously described, we determined the Ka values that range between (0.27 ( 0.05) 104 M1 and (4.0 ( 0.5) 104 M1 at pH 3.4; at pH 8.4 Ka values are between (14.5 ( 0.1) 104 M1and (390 ( 20) 104 M1 (Figure 3, Table 1). To conclude, we obtained the Ka values of concanavalin at pH 3.4, pH 6.8, and pH 8.4 by MS-based titration. Determining the Ka values allowed us to rank the strength of the binding between dimers of concanavalin A at three different pHs. We can also observe that the higher the pH, the higher the Ka value, that is, high pH favors the concanavalin A tetramer. Even though the absolute values of Ka were quite different depending on the chosen factor R (see below), ranking the binding strengths at the three pH values was always possible (i.e., Ka‑8.4> Ka‑6.8> Ka‑3.4), that is, relative binding affinities was determined independently from the factor R. Comparison of Ka Values Determined by MS and by ITC. After using nano-ESI-MS, we determined the Ka values for the dimertetramer equilibrium of concanavalin A in solution at three different pH values, we decided to compare the MS-based Ka values to those obtained by another biophysical approach. Isothermal titration calorimetry (ITC) is considered “the gold standard” to measure the solution phase Ka value, the enthalpy change (ΔH), the Gibbs free energy change (ΔG), and the entropy change (ΔS) of a binding interaction. To investigate the concanavalin A dimertetramer equilibrium, we utilized the ITC-based “dissociation model”, as previously described.2 We chose ITC because it allows us to determine the Kd in solution, utilizing the same buffer (100 mM NH4Ac) as used for MS experiments. Like MS, ITC is also a label-free technique and does not require any immobilization of biomolecules. Nevertheless, the use of ITC posed several problems. First, concanavalin A has a solubility of 200 μM, close to the sensitivity limit of the ITC system.2 Second, the total amount of concanavalin A required for an ITC experiment was 120 times larger than the one necessary for the MS-based titration at a specific pH. Third, the optimization of the ITC experiments was very time-consuming, because we needed to carefully establish the appropriate amount of concanavalin A to be injected, because of the limited sensitivity of the calorimetric apparatus in detecting the heat of dissociation. Fourthly, the mathematical model, which the Microcal software uses, 9256
dx.doi.org/10.1021/ac201576e |Anal. Chem. 2011, 83, 9251–9259
Analytical Chemistry has been designed for a monomerdimer equilibrium. We analyzed a dimertetramer equilibrium, and a residual amount of monomer could compromise the data fitting. Despite these problems, we obtained the Kd of the dimertetramer equilibrium at pH 6.8 by using eq 5 (see Experimental Section) to fit the ITC data (Figure 4). For simplicity we report the Ka value at pH 6.8, which is the inverse of Kd obtained by ITC (Table 1). The ITC-determined Ka value [(4.1 ( 1.8) 104 M1] was in agreement with the Ka obtained by MS-based titration for factor R ≈ 4. This means that it was possible to restrict the value of factor R with the aid of the ITC results. We also carried out ITC experiments to determine the Kd values of concanavalin A at pH 8.4 and pH 3.4. At the basic pH, we obtained a Ka value of (21.0 ( 12.0) 104 M1 (Figure 4). We did not succeed in the ITC analyses of concanavalin A at pH 3.4 The abundance of tetramer at this acidic pH was very small (Figure 2a and b), and the heat qi associated with the dissociation of low abundant tetramers during each injection of the complex into the cell was insufficient to be determined by the ITC apparatus. Overall, using a large amount of protein, we determined the Kd of concanavalin at pH 6.8 and pH 8.4, but not at pH 3.4. ITC is certainly not an ideal method to evaluate the strength of the interactions between concanavalin A dimers. If we compare the Ka values determined by MS-based titration (Ka_MS) and by ITC (Ka_ITC) (Table 1), the Ka_MS > Ka_ITC when the factor R is e1. If we consider Ka_ITC the most accurate value, the Ka_MS> Ka_ITC indicates that concentration of the tetramer at equilibrium [T] is overestimated or [D] is underestimated by the MS-based method. Taking into account ([T]/[D]) = (IT/ID)(RD/RT) (see also eq 3) we can observe that [T] is inversely proportional to the response factor of the tetramer RT. This is consistent with the fact that the best agreement between Ka_MS and Ka_ITC is when the factor R is between 1 and 6. This indicates that RT is larger than RD, meaning that the tetramer is ionized, transmitted, and detected better than the dimer. Many different events could contribute to response factors of protein complexes and the literature on this subject is still scarce.15,16 It is, for example, not clear whether larger proteins generally have higher response factors than smaller ones.29 We focused our attention to the ionization probability of concanavalin A, considering that the ion transmission and detection of our instrument have been optimized for ions with m/z between 2000 and 8000.16 Taking into account the charged residue model (CRM), the ionization probability of a protein (or a protein complex) depends on its protonated basic residues and surface activity. Kebarle and co-workers30 proposed that protonated basic amino acids could be part of the charge at the surface of a droplet. When several protonated residues are attached to the droplet surface, it is more likely that the protein (or the protein complex) is present in one of the offspring droplets because charges preferentially enter the offspring droplets. 29 In the case of concanavalin A, the z values of the tetramer (18+, 19+, 20+ and 21+) are larger than the ones of the dimer (13+, 14+, 15+ and 16+). The available basic residues, which are not involved in bound formation and free to be protonated, are more numerous for the concanavalin A tetramer (52 residues) than for the dimer (32 residues).21 Therefore, more charges and more available basic residues for the concanavalin A tetramer argue for a larger RT than RD. Also, Konermann and co-workers suggested that a high surface activity of a protein (or a protein complex) also could enhance the ionization probability when the CRM applies.28 They explain that small
ARTICLE
offspring droplets, formed at the most peripheral layer of a parent droplet, lead to the formation of gas-phase ions. Therefore, a highly surface active protein (or a protein complex) could exhibit an enhanced ionization probability. If we correlate surface activity and oligomeric state, the concanavalin A tetramer can have an higher activity relative than its dimer due to the increasing number of contacts with the ESI droplet surface.31 We also calculated the SES, considering the crystallographic structure of dimers and tetramers (1CON in the PDB database). We found that the SES of the tetramer (26263 Å2) is larger than that of the dimer (15892 Å2), contributing to a higher RT than RD. To conclude, a larger number of charges and of protonable sites, and a greater SES of the concanavalin A tetramer positively affects its response factor, explaining the better agreement of the Ka values determined by ITC and by MS, when the 1 < R < 6 (Table 1).
’ CONCLUSIONS Using nano-ESI-MS, we investigated the oligomeric states of the lectin concanavalin A and showed that the gas-phase signal intensity of dimers and tetramers depended on the solution protein concentration and on the buffer pH. As the solution concentration of concanavalin A increased, the equilibrium shifted toward the tetramer and the abundance of this oligomeric species in the gas phase rose. The tetramer was also favored when the solution pH was basic (i.e., pH 8.4), and this can be explained by considering the different types of interactions between two dimers forming a tetramer. 21 At acidic pH (i.e., pH 3.4) the dimer was the dominant species, although the tetramer was still detectable. We were able to determine the Ka values of the dimer tetramer equilibria using MS-based titration, taking into consideration the different probabilities of being ionised, transmitted and detected of the dimers and tetramers. We showed that the titration method is an excellent approach to establish the Ka values of proteinprotein interactions when different oligomeric species are in equilibrium. Independently from the factor R, we were able to rank the Ka values and therefore compare the strength of the interactions between dimers of concanavalin A at three different pHs (i.e., relative binding affinity, Ka‑8.4> Ka‑6.8> Ka‑3.4). We correlated the investigation of concanavalin A by MS with the analysis by ITC. Both MS and ITC allowed the determination of the Ka values in the same buffer (100 mM NH4Ac), both are a label-free and do not require any immobilization of biomolecules. Compared to ITC, MS required a much smaller amount of sample for each analysis: 0.13 mg of concanavalin A versus 15 mg for ITC. The large amount of material represents a key drawback of ITC, in particular when protein complexes, which are analyzed, are difficult to express or to purify in vivo. Another advantage of MS is the possibility of directly monitoring the presence of distinct oligomeric species at different pH values. On the contrary, using ITC the existence of dimers and of tetramers can be only monitored indirectly by the heat that is produced by the dissociation of the tetramers into dimers. The presence of monomers due to the dimer dissociation could affect the ITC data analysis. Furthermore, the low abundance of the tetramer prevented the determination of the Ka values of dimer-tetramer equilibrium at pH 3.4 by ITC. By correlating Ka_MS and Ka_ITC values, we restricted the factor R for concanavalin A. Compared to other published studies on proteinprotein binding affinities,15,16 the results of our 9257
dx.doi.org/10.1021/ac201576e |Anal. Chem. 2011, 83, 9251–9259
Analytical Chemistry
ARTICLE
investigation illustrate the influence of factor R on the accurate determination of Ka values, when quantitatively studying interactions between proteins by MS. Knowing the factor R allowed us to calculate the Ka at pH 3.4 using MS. This latter experiment was impossible by ITC alone because of its limited sensitivity. A foreseen use of our methodology is that a single ITC experiment could provide the factor R, and then many MS-based measurements can be carried out at different pH or for testing different mutants of a protein. Further studies will be necessary to determine electrospray response factors of proteinprotein interactions directly by MS. The possibility of obtaining response factors by MS, independently from determination of Ka values, has been proposed by Gabelica et al. for a receptor molecule M and ligands L forming M 3 L complexes.27 Overall, our investigation indicates that MS is an excellent method for determining binding strength of proteinprotein interactions within noncovalent complexes. The substantial advantages of MS over other techniques (e.g., ITC) are sensitivity and the ability to provide information on composition, stoichiometry and subunit interactions of protein complexes. Our method has wide application because a small amount of unlabeled sample is necessary for each analysis and the ESI-MS data of the protein complexes reflect the solution phase composition.32,33
’ APPENDIX I: DERIVATION OF EQ 4 The analytical concentration of dimer D0, determined by UV spectrophotometry, is equal to the equilibrium concentration of the dimer D added to twice the concentration of the tetramer T (eq 7). ½D0 ¼ ½D þ 2½T
ð7Þ
The areas under the peaks representing the dimers and the tetramers are called as ID and IT. The ratio of ID/IT was plotted versus the analytical concentration of the dimer D0 (Figure 1c and 3). Using the equations reported below, we determined the constant of association (Ka) of dimertetramer interaction at pH 3.4, pH 6.8, and pH 8.4 and the factor R, which is the ratio between the response factors of the corresponding analytes (eq 3). Ka ¼
1 ½T ¼ Kd ½D2
ð2Þ
RT RD
ð3Þ
I D ¼ R D ½D
ð8aÞ
I T ¼ R T ½T
ð8bÞ
½D I T R ¼ ½T I D
ð9aÞ
½T 1 IT ¼ ½D R ID
ð10Þ
R ¼
In eq 2, one can substitute [T]/[D] from eq 10 to obtain Ka ¼
1 IT 1 R I D ½D
ð11Þ
In eq 11, [D] can be expressed using eq 7 Ka ¼
1 IT R ID
1 ½T ½D0 1 2 ½D0
ð12Þ
In eq 12, 1/[D0] can be expressed using eq 7 Ka ¼
1 IT R ID
0
1
B B 1 ½D0 B1 2 @ ½D þ ½T
1
2
ð13Þ
C C C A
In eq 13, [D]/[T] can be expressed using eq 10 ID R þ 2 1 IT IT Ka ¼ R I D ½D R ID 0 IT
ð14Þ
Equation 14 can be expressed as quadratic equation 2 ID ID K a ½D0 R R 2 ¼ 0 IT IT 2
ð15Þ
Equation 15 can be solved for the variable (ID/IT) pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ID 1 þ 1 þ 8K a D0 ¼ 2RK a D0 IT
ð4Þ
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.
’ ACKNOWLEDGMENT We thank Dr. Vladimir Torbeev (Laboratory of Organic Chemistry, ETH Zurich, Switzerland), Dr. Daniel Barsky (UC Berkeley, USA) and Dr Helena Hernandez (University of Oxford, UK) for useful discussions. We are grateful to Dr. Lumi Damian (GE Healthcare, MicroCal Products Group), Mr Michael Schneider and Mr Peter Kaelin (both at the Microlaboratory, Laboratory of Organic Chemistry, ETH Zurich) for their assistance during the isothermal titration calorimetry experiments. This scientific work was financially supported by the Swiss National Science Foundation (grant no. 200020-111831/1). ’ REFERENCES (1) Robinson, C. V.; Sali, A.; Baumeister, W. Nature 2007, 450, 973–982. (2) Velazquez-Campoy, A.; Leavitt, S. A.; Freire, E. Methods Mol. Biol. 2004, 261, 35–54. (3) Burrows, S. D.; Doyle, M. L.; Murphy, K. P.; Franklin, S. G.; White, J. R.; Brooks, I.; McNulty, D. E.; Scott, M. O.; Knutson, J. R.; Porter, D.; et al. Biochemistry 1994, 33, 12741–12745. (4) Chen, S.; Chen, L.; Tan, J.; Chen, J.; Du, L.; Sun, T.; Shen, J.; Chen, K.; Jiang, H.; Shen, X. J. Biol. Chem. 2005, 280, 164–173. (5) Sakurai, K.; Oobatake, M.; Goto, Y. Protein Sci. 2001, 10, 2325– 2335. 9258
dx.doi.org/10.1021/ac201576e |Anal. Chem. 2011, 83, 9251–9259
Analytical Chemistry
ARTICLE
(6) Boeri Erba, E.; Zenobi, R. Annu. Rep. Prog. Chem., Sect. C 2011, 107. (7) Chitta, R. K.; Rempel, D. L.; Gross, M. L. J. Am. Soc. Mass Spectrom. 2005, 16, 1031–1038. (8) Daniel, J. M.; McCombie, G.; Wendt, S.; Zenobi, R. J. Am. Soc. Mass Spectrom. 2003, 14, 442–448. (9) Jecklin, M. C.; Schauer, S.; Dumelin, C. E.; Zenobi, R. J. Mol. Recognit. 2009, 22, 319–329. (10) Wilcox, J. M.; Rempel, D. L.; Gross, M. L. Anal. Chem. 2008, 80, 2365–2371. (11) Annis, D. A.; Nazef, N.; Chuang, C. C.; Scott, M. P.; Nash, H. M. J. Am. Chem. Soc. 2004, 126, 15495–15503. (12) El-Hawiet, A.; Kitova, E. N.; Liu, L.; Klassen, J. S. J. Am. Soc. Mass Spectrom. 2010, 21, 1893–1899. (13) Wortmann, A.; Jecklin, M. C.; Touboul, D.; Badertscher, M.; Zenobi, R. J. Mass Spectrom. 2008, 43, 600–608. (14) Gabelica, V.; Galic, N.; Rosu, F.; Houssier, C.; De Pauw, E. J. Mass Spectrom. 2003, 38, 491–501. (15) Ayed, A.; Krutchinsky, A. N.; Ens, W.; Standing, K. G.; Duckworth, H. W. Rapid Commun. Mass Spectrom. 1998, 12, 339–344. (16) Liu, J.; Konermann, L. J. Am. Soc. Mass Spectrom. 2011, 22, 408–417. (17) Agrawal, B. B. L.; Goldstein, I. J. Arch. Biochem. Biophys. 1968, 124, 218–229. (18) Lu, Q.; Lin, H.; Ge, S.; Luo, S.; Cai, Q.; Grimes, C. A. Anal. Chem. 2009, 81, 5846–5850. (19) Wilchek, M.; Chaiken, I. Methods Mol. Biol. 2000, 147, 1–6. (20) Lei, H. Y.; Chang, C. P. J. Biomed. Sci. 2009, 16, 10. (21) Naismith, J. H.; Habash, J.; Harrop, S.; Helliwell, J. R.; Hunter, W. N.; Wan, T. C.; Weisgerber, S.; Kalb, A. J.; Yariv, J. Acta Crystallogr. Sect. D Biol. Crystallogr. 1993, 49, 561–571. (22) Senear, D. F.; Teller, D. C. Biochemistry 1981, 20, 3076–3083. (23) Boeri Erba, E.; Bergatto, E.; Cabodi, S.; Silengo, L.; Tarone, G.; Defilippi, P.; Jensen, O. N. Mol. Cell. Proteomics 2005, 4, 1107–1121. (24) Sobott, F.; Hernandez, H.; McCammon, M. G.; Tito, M. A.; Robinson, C. V. Anal. Chem. 2002, 74, 1402–1407. (25) van den Heuvel, R. H.; van Duijn, E.; Mazon, H.; Synowsky, S. A.; Lorenzen, K.; Versluis, C.; Brouns, S. J.; Langridge, D.; van der Oost, J.; Hoyes, J.; Heck, A. J. Anal. Chem. 2006, 78, 7473–7483. (26) Reger, D. L.; Goode, S. R.; Ball, D. W. Chemistry: Principles and Practice; Brooks/Cole Publishing Company: Pacific Grove, CA, 2009. (27) Gabelica, V.; Rosu, F.; De Pauw, E. Anal. Chem. 2009, 81, 6708–6715. (28) Kuprowski, M. C.; Konermann, L. Anal. Chem. 2007, 79, 2499– 2506. (29) Peschke, M.; Verkerk, U. H.; Kebarle, P. J. Am. Soc. Mass Spectrom. 2004, 15, 1424–1434. (30) Verkerk, U. H.; Peschke, M.; Kebarle, P. J. Mass Spectrom. 2003, 38, 618–631. (31) Magdassi, S.; Kamyshny, A. Surface Activity and Functional Properties of Proteins Surface Activity of Proteins: Chemical and Physicochemical Modifications; Marcel Dekker, Inc.: New York, NY, 1996. (32) Benesch, J. L.; Ruotolo, B. T.; Simmons, D. A.; Barrera, N. P.; Morgner, N.; Wang, L.; Saibil, H. R.; Robinson, C. V. J. Struct. Biol. 2010, 172, 161–168. (33) Ruotolo, B. T.; Giles, K.; Campuzano, I.; Sandercock, A. M.; Bateman, R. H.; Robinson, C. V. Science 2005, 310, 1658–1661.
9259
dx.doi.org/10.1021/ac201576e |Anal. Chem. 2011, 83, 9251–9259
