Quantifying Protein Interactions with Isomeric Carbohydrate Ligands

of Chemistry, University of Alberta, Edmonton, Alberta, Canada T6G 2G2 ... Please contact your librarian to recommend that your institution subscr...
0 downloads 0 Views 665KB Size
Download PDF
Editors' Highlight pubs.acs.org/ac

Quantifying Protein Interactions with Isomeric Carbohydrate Ligands Using a Catch and Release Electrospray Ionization-Mass Spectrometry Assay Amr El-Hawiet, Elena N. Kitova, and John S. Klassen* Alberta Glycomics Centre and Department of Chemistry, University of Alberta, Edmonton, Alberta, Canada T6G 2G2 S Supporting Information *

ABSTRACT: The application of a catch-and-release electrospray ionization mass spectrometry (CaR-ESI-MS) assay to quantify interactions between proteins and isomeric carbohydrate ligands is described. Absolute affinities for each ligand are determined from the abundance ratio of ligand-bound to free protein measured directly by ESI-MS and the relative abundances of the individual isomeric ligands, which are established by releasing the ligands, in their deprotonated form, from the protein using collision-induced dissociation (CID) and subjecting them to ion mobility separation (IMS) or another stage of CID to fragment the ions. Using Gaussian functions to represent the contributions of individual ligands to the arrival time distributions (ATDs) measured by IMS, the relative abundance of each ligand bound to the protein can be established. A modification of this method, suitable for cases where nonspecific ligand-protein binding occurs during the ESI process, is also described. In cases where the ATDs are not sufficiently different to distinguish the isomeric ligands, CID can establish the relative abundance of each ligand bound to the protein from the relative abundance of the resulting fragment ions. The implementation and reliability of the CaR-ESI-MS assay for the analysis of isomeric carbohydrate ligands is demonstrated using three carbohydrate-binding proteins, a single chain antibody, an antigen binding fragment, and a fragment of a bacterial toxin, and their interactions with isomeric carbohydrate ligands with affinities ranging from 103 to 105 M−1.

I

binding stoichiometry. Consequently, other binding assays must be used to establish the binding stoichiometry and to quantify the affinity of identified interactions. Recently, electrospray ionization mass spectrometry (ESIMS), implemented using a catch-and-release (CaR) strategy, has emerged as a promising technique for screening libraries of carbohydrates against target proteins.13−15 The CaR-ESI-MS assay has a number of attractive features, including its simplicity, speed, low sample consumption, and the unique ability to directly probe binding stoichiometry and affinity.13 The assay involves direct ESI-MS analysis of the target protein(s) in the presence of a mixture of carbohydrates to detect specific protein−ligand complexes. In many instances, the identity of ligands (“caught” by the protein) can be found from the molecular weight (MW) of the corresponding protein−ligand complex, as determined from the ESI mass spectrum. In cases where MW cannot be accurately determined (due to the size or heterogeneity of the protein) or when dealing with isomeric ligands, the ligands are “released” as ions from the protein using collision-induced dissociation (CID), followed by accurate mass analysis alone or in combination with ion mobility separation (IMS) or another stage of CID.13 The CaR-ESI-MS assay was recently used to identify and

t is widely recognized that noncovalent interactions between proteins and carbohydrates are involved in many critical biochemical processes, including molecular recognition, catalysis, and signaling.1 Consequently, the development of new analytical methods that can accelerate the discovery and characterization of biologically relevant protein−carbohydrate interactions represents an active area of research. Currently, glycan microarrays, which utilize immobilized oligosaccharides or glycans from natural sources or produced using enzymatic and chemoenzymatic strategies, represent the dominant technology for screening carbohydrate libraries against target proteins (of mammalian, bacterial, or plant origin).2−10 The development of carbohydrate microarrays has catalyzed the dramatic increase in the number of known protein− carbohydrate interactions, aided in the discovery of new glycosyltransferases, and facilitated the characterization of the glycan specificities of many viral and bacterial pathogens.6,9 The strengths of glycan microarrays include speed of screening (hundreds of interactions can be probed within a few days), low protein and glycan consumption (1−100 μg of protein and oligosaccharide10), and tolerance to the presence of impurities in the protein sample.2,6 However, complexes with high off rates can be lost during the washing step, resulting in false negatives.11 Additionally, glycan array screening generally provides only qualitative binding data, and at best, it can rank the relative affinities when implemented using a titration strategy.12 Moreover, the technique provides no insight into © 2013 American Chemical Society

Received: May 31, 2013 Accepted: July 29, 2013 Published: July 29, 2013 7637

dx.doi.org/10.1021/ac401627t | Anal. Chem. 2013, 85, 7637−7644

Analytical Chemistry

Editors' Highlight

β-D-Gal-(1→4)-β-D-GlcNAc-(1→3)-β-D-Gal-(1→4)-β-D-Glc (L7, MW 707.63 Da) were purchased from IsoSep AB (Sweden). The structure of each ligand is shown in Figure S1 in the Supporting Information. Stock solutions of each carbohydrate were prepared by dissolving a known mass of carbohydrate, determined gravimetrically, in ultrafiltered water (Milli-Q, Millipore) to give a final concentration of 1 mM. These were stored at −20 °C until needed. Mass Spectrometry. All experiments were carried out using a Synapt G2 ESI quadrupole-ion mobility separationtime-of-flight (Q-IMS-TOF) mass spectrometer (Waters, Manchester, U.K.), equipped with a nanoflow ESI (nanoESI) source. Mass spectra were obtained in negative ion mode using cesium iodide (concentration 30 ng μL−1) for calibration. To perform nanoESI, tips were produced from borosilicate capillaries (1.0 mm o.d., 0.68 mm i.d.), which were pulled to ∼5 μm using a P-97 micropipet puller (Sutter Instruments, Novato, CA). A platinum wire was inserted into the nanoESI tip and a capillary voltage of 1.0−1.3 kV was applied. A cone voltage of 20−60 V was used, and the source block temperature was maintained at 70 °C. Other important voltages for ion transmission, such as the injection voltages into the trap, IMS, and transfer ion guides, were maintained at 5 V, 45 V, and 2 V, respectively. To release the ligand from the complex trap voltages ranging from 20 to 120 V were used. Argon was used in the trap and transfer ion guides at a pressure of 2.22 × 10−2 mbar and 3.36 × 10−2 mbar, respectively. The helium chamber preceding the traveling wave ion mobility separation (TWIMS) device was maintained at 7.72 mbar. All IMS measurements were carried out using N2 as the mobility gas, at a pressure of 3.41 mbar. Data acquisition and processing were carried out using MassLynx (v4.1). CID of the released ligands was carried out in the transfer region at voltages ranging from 12 to 20 V. To determine the relative abundance of protein-bound isomeric ligands, deprotonated ions of the corresponding protein−carbohydrate complexes were isolated using the quadrupole mass filter and then subjected to CID in the trap ion guide to release the ligands. Two different approaches to quantify isomeric ligands were used. In the first approach, the ATDs of the released ligands were measured by IMS. Drift times, corresponding to the maxima on the ATDs, were used to confirm identity of the released ligands. ATDs of the deprotonated ions of the individual ligands measured using the same IMS parameters served as references.13 To find the ATD peak areas corresponding to a particular isomeric ligand, the IMS mass lists (i.e., ion intensity versus drift time (in milliseconds)) of the individual and released ligands were exported into Igor Pro 6.2 (WaveMetrics Inc.) and the data analyzed with the Multi-Peak Fit package. A number of standard functions provided by Igor Pro were tested, and the Gaussian function was found to give the best fit to the ATDs measured for the individual oligosaccharide ions. Therefore, Gaussian functions were used to analyze the ATDs measured for released ligands. As described in detail in the following section, the contribution of each released ligand to the measured ATD was used to establish the relative abundances of the bound ligands and to quantify their affinities for the protein. In cases where the ATDs of individual isomeric ligands are not sufficiently different, the released ligands were fragmented using another stage of CID in the transfer region. The CID mass spectra of the deprotonated carbohydrates produced directly from solution served as references. The relative abundances of the isomeric ligands were established

quantify, in a single measurement, multiple moderate affinity ligands within a well-defined library of >200 carbohydrates.13 The assay has also been used to screen glycan libraries, prepared from natural sources, against carbohydrate-binding proteins. For example, Cederkvist et al. used CaR-ESI-MS to screen heterochitooligosaccharides, obtained by enzymatic hydrolysis of chitosan, against chitinolytic enzyme Chitinase B,14 while Kaltashov and co-workers used this approach to screen heterogeneous oligoheparin mixtures against antithrombin III.15 The aforementioned results are encouraging and suggest that the CaR-ESI-MS assay can serve as a complementary approach to glycan array technology for carbohydrate library screening. Furthermore, the assay can, in principle, provide direct insight into binding stoichiometry and affinity, all in a single measurement. However, the presence of structural isomers in the library can complicate the identification of specific ligands and the determination of absolute affinities. While it is possible to distinguish isomeric ligands binding to the same protein by subjecting them to IMS following their release,16−24 it has not yet been established whether absolute affinities can be reliably determined using this approach. The goal of this study was to explore the feasibility of using the CaR-ESI-MS assay to not only detect the binding of isomeric carbohydrate ligands to target proteins but also to quantify the interactions based on the relative abundances of the released ligands. With this objective in mind, control experiments were performed on three different carbohydratebinding proteins and their interactions with isomeric ligands, ranging in size from tri- to pentasaccharide and with affinities of between 103 and 105 M−1. Notably, it is shown that “deconvolution” of the IMS arrival time distributions (ATDs) for the released structural isomers can lead to absolute affinities that are in good agreement with reported values. In instances where the ATDs are not sufficiently different to distinguish isomeric ligands, an alternative approach, based on the relative abundances of fragment ions produced by CID, to quantify the relative abundance of each ligand bound to protein is shown to be effective.

■

EXPERIMENTAL SECTION Proteins and Ligands. An antigen binding fragment (Fab, MW 48 263 Da) of the monoclonal antibody (mAb) CS-35 and a single chain fragment (scFv, MW 26 539 Da) of the mAb Se155-4 were produced and purified as described previously.25,26 A fragment (B3) of the carboxy-terminus of the toxin TcdB from Clostridium dif f icile strain 630 (TcdB-B3, MW 30 241 Da) was a gift of Prof. K. Ng (University of Calgary). Each protein was concentrated and dialyzed against aqueous 50 mM ammonium acetate solution using microconcentrators (Millipore Corp., Bedford, MA) with a MW cutoff of 10 kDa and stored at −20 °C, if not used immediately. The carbohydrate ligands, β-D-Araf-(1→2)-α-D-Araf-(1→5)-[α-DAraf-(1→3)]-α-D-Araf-(1→5)-α-D-Araf-OCH 3 (L1, MW 692.61 Da), β-D-Araf-(1→2)-α-D-Araf-(1→3)-[α-D-Araf-(1→ 5)]-α-D-Araf-(1→5)-α-D-Araf-OCH3, (L2, MW 692.61 Da) were gifts from Prof. T. Lowary (University of Alberta). α-DTalp-(1→2)-[α-D-Abep-(1→3)]-α-D-Manp-OCH3 (L3, MW 486.46 Da), α-D-Glcp-(1→2)-[α-D-Abep-(1→3)]-α-D-ManpOCH3 (L4, MW 486.46 Da) and α-D-Galp-(1→2)-[α-D-Abep(1→3)]-α-D-Talp-OCH3, (L5, MW 486.46 Da) were gifts from Prof. D. Bundle (University of Alberta). β-D-Gal-(1→3)-β-DGlcNAc-(1→3)-β-D-Gal-(1→4)-β-D-Glc (L6, 707.63 Da) and 7638

dx.doi.org/10.1021/ac401627t | Anal. Chem. 2013, 85, 7637−7644

Analytical Chemistry

Editors' Highlight

from a comparison of the relative abundances of fragment ions produced from the mixture of released ligands and from the individual carbohydrate ions. Determination of Ka for Isomeric Ligands. With the direct ESI-MS assay, the Ka value for a given protein−ligand interaction is determined from the ratio (R) of the total abundance (Ab) of ligand-bound (PL) and free protein (P) ions measured for a solution of known initial concentration of protein ([P]o) and ligand ([L]o). For a 1:1 PL complex (eq 1), Ka can be calculated using eq 2.

where SATD,i is the area of the ATD corresponding to a given Li ion. It follows that [PLi] can be calculated from eq 8,

(1)

(9)

P + L ⇌ PL

[PLi] = fi

i

[L]o −

[Li] = [Li]o −

(2)

K a,Li =

∑i Ab(PLi) Ab(P)

(4a)

where Ab(PLi) is the total abundance of PLi ions (as determined from the mass spectrum). Rapp can also be expressed in terms of concentrations, eq 4b,

Frapp =

[P]

(4b)

where [PLi] is the concentration of the corresponding PLi species in solution. From mass balance considerations (eq 5), [P] and ∑i[PLi] can be determined from eqs 6a and 6b, respectively,

∑ [PLi] i

[P] =

[P]o 1 + R app

∑ [PLi] = i

1 + R app

[P]o

(5)

(6b)

The concentration of individual PLi species can be determined from the fractional abundance (or fractional concentration) of each PLi complex (f i), which can be established from the contribution that each released Li ion makes to the total area of the ATD, eq 7, fi =

SATD, i ∑i SATD, i

=

[PLi] ∑i [PLi]

(1 + R app)

[P]o

fi R app [PLi] = f R app [P][Li] [Li]o − 1 +i R [P]o

(10)

∑ fi Fri

(11)

f1 = (Frapp − Fr2)/(Fr1 − Fr2)

(12a)

f2 = 1 − f1

(12b)

Additional fragment ions must be considered in situations where three or more isomeric ligands are present. Another important underlying assumption in the application of the CaR-ESI-MS assay is that all of the detected PLi ions originate from complexes that were present in bulk solution. However, as has been discussed in detail elsewhere, nonspecific binding between carbohydrates and proteins during the ESI process can produce false positives.27 In such cases, the relative abundances of the gaseous PLi ions measured by ESI-MS will not accurately reflect solution composition. It was shown previously that the reference protein method, which involves the addition of a noninteracting reference protein (Pref) to solution, can be used to quantitatively correct ESI mass spectra for the occurrence of nonspecific protein−ligand binding.28 The method draws on the statistical nature of the nonspecific binding process and the fact that nonspecific ligand binding uniformly affects all proteins present in the ESI droplets (in a given experiment). In the present study, the reference protein method was adapted to allow for the correction of ESI mass

(6a)

R app

(8)

where Fri is the abundance ratio of fragment and precursor ions for a given Li. In cases involving two isomeric ligands, Frapp for a single, common, fragment ion is sufficient to establish the corresponding f i values (i.e., f1 and f 2), eqs 12a and 12b,

∑i [PLi]

[P]o = [P] + [PL1] + ... + [PLi] = [P] +

[P]o

It is important to stress that underlying this approach is the assumption that the efficiency by which the ligands are released from the complex by CID is identical for all isomeric ligands, independent of their structures (and their solution affinities).13 As described in more detail below, the experimental data acquired in the present study provide support for the validity of this assumption. In cases where the released isomeric ligand ions cannot be distinguished by IMS (due to the similarity in collision cross sections), an alternative approach, which is based on relative abundances of fragment ions produced by CID, may be used to calculate the corresponding f i values. In general, the contribution of each isomer to the measured abundance of a common fragment ion (reported as the ratio of the abundances of the fragment and precursor ions, Frapp) is given by eq 11.

(3)

An underlying assumption in this approach is that PL and P have similar ionization and detection efficiencies (i.e., similar ESI-MS response factors) such that the abundance ratio of gas phase ions is equal to the concentration ratio in solution. This assumption has been shown to be valid in cases where L is small compared to P, such that P and PL are similar in size and surface properties.25 When two or more isomeric ligands (e.g., L1, L2, ...Li) are bound to P, the apparent (measured) abundance ratio (Rapp) is described by eq 4a,

R app =

fi R app

app

Ab(PL) [PL] = Ab(P) [P]

R app =

(1 + R app)

Finally, Ka,Li, the association constant for a given PLi complex, can be calculated from eq 10.

R [P]o 1+R

where R is given by eq 3, R=

fi R app

and the concentration of each ligand can be found from mass balance considerations, eq 9,

R

Ka =

∑ [PLi] =

(7) 7639

dx.doi.org/10.1021/ac401627t | Anal. Chem. 2013, 85, 7637−7644

Analytical Chemistry

Editors' Highlight

Affinities of CS-35 Fab for L1 and L2. The application of the CaR-ESI-MS assay to detect the specific interactions between the Fab of CS-35 and two isomeric pentasaccharide ligands (L1, Ka = (6.3 ± 0.7) × 103 M−1 and L2, Ka = (9.9 ± 0.5) × 104 M−1),25 which were present in a mixture of 203 carbohydrates, was recently reported.13 It was shown that, following their release from the protein, the deprotonated L1 and L2 ions could be partially resolved by IMS. Because the library components were at equimolar concentrations, it was possible, from the contribution of each ligand to the ATDs, to rank the affinities of L1 and L2. However, at the time, it was not established whether the relative abundances of released ligands quantitatively reflected the relative abundances of bound ligands. To answer this question, the CaR-ESI-MS assay was applied to solutions of CS-35 Fab and L1 and L2 at three different concentrations (ranging from 10 to 15 μM for L1 and from 1 to 10 μM for L2). A representative ESI mass spectrum acquired in negative ion mode for an aqueous ammonium acetate (10 mM) solution of CS-35 Fab (4.5 μM), L1 (12.5 μM) and L2 (2.5 μM) is shown in Figure S2a in the Supporting Information. As discussed previously, papain digestion of the CS-35 mAb produces four Fab isoforms (labeled as P, P′, P″ and P‴ in Figure S2a in the Supporting Information).25 Ions corresponding to both unbound and ligand-bound Fab isoforms were detected (at charge states −12 to −14). The abundance ratio of ligand-bound to free protein, i.e., Rapp, determined directly from the ESI mass spectrum, was found to be 0.24 ± 0.01. The (P + L)13− ions, corresponding to the most abundant Fab isoform, were isolated using the quadrupole mass filter and then subjected to CID (trap voltage 120 V) to release the ligands. Under these conditions, approximately 99% of the complex ions were dissociated, with the ligands (L1 and L2) released in their deprotonated form. No covalent fragmentation of the ligand or protein ions was observed. The deprotonated ligands were subjected to IMS using the following parameters: IMS gas pressure, 3.41 mbar; wave velocity, 350 m s−1; and wave height, 15 V. Inspection of the ATD profiles of the released ligands reveals the presence of two features, with maxima at approximately 5.80 and at 6.20 ms (Figure 1c). These features are consistent with the maxima observed in the ATDs measured for the individual L1 and L2 ions (Figure 1a,e). Following the procedure described in the Experimental Section, the fractional abundances of the released ligands were found to be 0.38 ± 0.07 (f L1) and 0.62 ± 0.05 ( f L2), and the corresponding Ka values for L1 and L2 were calculated to be (7.5 ± 0.8) × 103 and (7.6 ± 0.7) × 104 M−1, respectively (Table 1). Shown in Figure 1b,d are ATDs measured at two other ligand concentrations. It can be seen that the relative abundances of released L1 and L2 vary, as expected, with the initial ligand concentration. The Ka values determined for all three sets of ligand concentrations are similar, within a factor of 2 (Table 1), and the average values of (8.8 ± 0.8) × 103 M−1 and (1.1 ± 0.1) × 105 M−1 for L1 and L2, respectively, are in good agreement with the reported affinities.25 As noted above, underlying the use of the CaR-ESI-MS method to quantify affinities for isomeric ligands is the assumption of uniform “release efficiency” for the isomeric ligands. While this is necessarily the case under conditions when the complex is fully dissociated (to protein and ligand), it may not be the case under CID conditions that lead to only partial dissociation of the complex ions. To establish whether the relative abundances of the released ligands and,

spectra for nonspecific binding of isomeric ligands to a common protein. In cases where both specific and nonspecific binding contribute to the abundances of the PLi species, the apparent (measured) abundance ratio (Rapp*) is given by eq 13.

R app* = R sp + R ns

(13)

where Rns is the ratio of the total abundance of nonspecific PLi,ns complexes (as determined from the abundances of PrefLi species) to free P, eq 14a, ∑i Ab(PLi ,ns)

R ns =

=

Ab(P)

∑i Ab(Pref Li) Ab(Pref )

(14a)

and Rsp is the ratio of the total abundance of specific PLi,sp complexes to free P, eq 14b, ∑i Ab(PLi ,sp)

R sp =

Ab(P)

(14b)

The fractional abundance (f i,ns) of a given PrefLi complex can be obtained from the contribution of individual Li ions to the ATDs measured for the mixtures of ligands released from the PrefLi ions, eq 15, SATD, i ,ns

fi ,ns =

∑i SATD, i ,ns

(15)

The fractions of the PLi,sp and PLi,ns species present in a given PLi complex (i.e., Fsp and Fns, respectively) can be represented by eqs 16a and 16b,

Fsp =

Fns =

R sp R sp + R ns

(16a)

R ns R sp + R ns

(16b)

It follows that the apparent (measured) fractional abundance of a given PLi complex ( f i*) can be described by eq 17, f i* = fi ,sp

R sp R sp + R ns

+ fi ,ns

R ns R sp + R ns

(17)

where f *i can be calculated using the SATD,i values measured for the Li ions after release from the PLi ions, eq 18, f i* =

SATD, i ∑i SATD, i

(18)

Once f i,sp and Rsp are known, Ka,Li can be calculated using eq 19, K a,Li =

■

fi ,sp R sp [Li]o −

fi,sp R sp 1 + R sp

[P]o

(19)

RESULTS AND DISCUSSION To test the reliability of the CaR-ESI-MS approaches described above for quantifying the affinities of isomeric carbohydrate ligands for a common protein, control experiments were performed on three carbohydrate-binding proteins, the Fab of mAb CS-35, the scFv of mAb Se155-4, and fragment B3 of toxin B from Clostridium dif f icile, and their interactions with isomeric ligands with affinities ranging from 103 to 105 M−1. The results obtained for each protein are described below. 7640

dx.doi.org/10.1021/ac401627t | Anal. Chem. 2013, 85, 7637−7644

Analytical Chemistry

Editors' Highlight

determined, using the direct ESI-MS assay, to be (4.8 ± 0.6) × 104, (5.4 ± 0.9) × 104, and (4.3 ± 0.1) × 104 M−1, respectively.13 Shown in Figure S2b in the Supporting Information is a mass spectrum acquired in negative ion mode for an aqueous ammonium acetate (10 mM) solution of scFv, L3 and L4 (5 μM each). Ions corresponding to both unbound and ligand-bound scFv were detected at charge states 8− and 9−; the Rapp value was found to be 0.46. Using optimized IMS parameters (IMS gas pressure, 3.41 mbar; wave velocity, 600 m s−1; and wave height, 30 V), the ATD was measured for the ligands released from the (scFv + L)9− ions (Figure 2c). Two features are evident in the ATD, with maxima

Figure 1. ATDs measured for the deprotonated oligosaccharides L1 (a) and L2 (e). ATD measured for deprotonated L1 and L2 following their release from the (Fab + L)13− ions (where L = L1 or L2) produced by ESI from solutions of Fab CS35 (4.5 μM) and (b) L1 (15 μM) and L2 (1.3 μM), (c) L1 (12.5 μM) and L2 (2.5 μM), and (d) L1 and L2 (10 μM each). Dashed curves correspond to calculated ATD based on the fractional abundance of L1 and L2 of 0.68 and 0.32 (b), 0.37 and 0.63 (c), and 0.14 and 0.86 (d), respectively.

Table 1. Association Constants (Ka) for CS-35 Fab Binding to L1 (Ka,L1) and L2 (Ka,L2) Measured in Aqueous Ammonium Acetate (pH 7 and 23 °C) at Different Initial Ligand Concentrations Using the CaR-ESI-MS Assaya [CS-35 Fab]o (μM) 4.5 4.5 4.5 a

[L1]o (μM) 15.0 12.5 10.0

[L2]o (μM) 1.3 2.5 10.0 average

Ka,L1 (M−1) (1.1 (7.5 (8.5 (8.8

± ± ± ±

0.4) 0.8) 0.4) 0.8)

× × × ×

Ka,L2 (M−1) 4

10 103 103 103

(1.7 (7.6 (9.1 (1.1

± ± ± ±

0.6) 0.7) 0.5) 0.1)

× × × ×

105 104 104 105

Errors correspond to one standard deviation.

consequently, Ka values, are sensitive to the CID conditions (and degree of dissociation), measurements were carried out at four different trap voltages, ranging from 60 to 120 V. Under these conditions, the extent of complex dissociation varied from 40% to 99%. As can be seen from Table S1 in the Supporting Information, the trap voltage (collision energy) has no significant effect on the measured Ka values. These results are qualitatively consistent with those of a recent CID study of structurally related protein−ligand complexes, which demonstrated very similar dissociation profiles (% dissociation versus collision energy) for complexes composed of isomeric carbohydrate ligands.29 Affinities of Se155-4 scFv for L3, L4, and L5. As a further test of the reliability of CaR-ESI-MS approach for quantifying the affinities of isomeric carbohydrate ligands, measurements were performed on solutions of Se155-4 scFv (5 μM) and isomeric trisaccharide ligands (L3, L4, and L5). The Ka values for L3, L4, and L5 binding to scFv were previously

Figure 2. ATDs measured for the deprotonated oligosaccharides L3 (a) and L4 (e). ATD measured for the deprotonated ligands L3 and L4, following their release from the (scFv + L)9− ions (where L = L3 or L4) produced by ESI from solutions of scFv (5 μM) and (b) L3 (5 μM) and L4 (2.5 μM), (c) L3 and L4 (5 μM each) and (d) L3 (2.5 μM) and L4 (5 μM). Dashed curves correspond to calculated ATD based on the fractional abundance of L3 and L4 of 0.63 and 0.37 (b), 0.55 and 0.45 (c), and 0.36 and 0.64 (d), respectively.

at 7.15 and 7.48 ms, consistent with the ATDs measured for the individual L3 and L4 ions (Figure 2a,e). The fractional abundances of the released L3 and L4 ions were found to be f L3 = 0.54 ± 0.03 and f L4 = 0.46 ± 0.03, respectively, which lead to Ka values of (5.3 ± 0.4) × 104 M−1 and (5.6 ± 0.5) × 104 M−1, respectively (Table 2). Measurements performed at two other sets of ligand concentrations (Figure 2b,d) yielded similar results (Table 2). The average Ka values, (5.3 ± 0.6) × 104 M−1 7641

dx.doi.org/10.1021/ac401627t | Anal. Chem. 2013, 85, 7637−7644

Analytical Chemistry

Editors' Highlight

305.1), the fractional abundances of the released L4 and L5 ions were found to be f L4 = 0.29 ± 0.01 and f L5 = 0.71 ± 0.01, respectively. The corresponding Ka values are (3.7 ± 0.1) × 104 M−1 and (5.8 ± 0.1) × 104 M−1, respectively. Similar results were obtained when other fragments ions were used. For example, using the m/z 323 fragment ion gave Ka values of (6.1 ± 1.0) × 104 M−1 (L4) and (4.5 ± 0.9) × 104 M−1 (L5). Notably, these values are in reasonable agreement with the values determined using the direct ESI-MS assay. Affinities of TcdB-B3 for L6 and L7. The interaction between the B3 fragment of toxin B of Clostridium dif f icile (TcdB-B3) and two human milk oligosaccharides (L6 and L7) served to demonstrate the application of the CaR-ESI-MS strategy to quantify isomeric ligand binding in cases where nonspecific carbohydrate−protein binding occurs during the ESI process.13 The affinities of L6 and L7 for TcdB-B3, which were measured using the direct ESI-MS assay, are (7.8 ± 1.9) × 103 M−1 and (5.7 ± 1.4) × 103 M−1, respectively. For these measurements, a Pref was added to the ESI solutions to correct the mass spectra for the occurrence of nonspecific binding.28 Shown in Figure S2c in the Supporting Information is a representative ESI mass spectrum acquired in negative ion mode for an aqueous ammonium acetate (100 mM) solution of TcdB-B3 (8.2 μM), L6, and L7 (4 μM each) and Pref (15 μM). Ions corresponding to both unbound and ligand-bound TcdBB3 ions were detected at charge states −10 and −11. Free and ligand-bound Pref ions at charge states −5 and −6 were also detected, indicating the occurrence of nonspecific ligand binding. The Rapp* and Rns values were 0.12 and 0.08, respectively; correcting for nonspecific ligand binding using the Pref method gave an Rsp value of 0.04.28 Using optimized IMS parameters (IMS gas pressure, 3.41 mbar; wave velocity, 600 m s−1; wave height, 30 V), ATDs were measured for the L6 and L7 ions released from the (Pref + L)6− (Figure 4a), (TcdB-

Table 2. Association Constants (Ka) for Se155-4 scFv Binding to L3 (Ka,L3) and L4 (Ka,L4) Measured in Aqueous Ammonium Acetate (pH 7 and 23 °C) at Different Initial Ligand Concentrations Using the CaR-ESI-MS Assaya [scFv]o (μM) 5.0 5.0 5.0 a

[L3]o (μM) 5.0 5.0 2.5

[L4]o (μM) 2.5 5.0 5.0 average

Ka,L3 (M−1) (5.6 (5.3 (4.9 (5.3

± ± ± ±

0.3) 0.4) 0.6) 0.6)

× × × ×

Ka,L4 (M−1) 4

10 104 104 104

(5.4 (5.6 (6.6 (5.8

± ± ± ±

0.5) 0.5) 0.3) 0.5)

× × × ×

104 104 104 104

Error corresponds to one standard deviation.

and (5.8 ± 0.5) × 104 M−1, agree very well with the values obtained using the direct ESI-MS assay.13 Notably, the magnitude of the Ka values were also found to be independent of the trap voltage used (Table S2 in the Supporting Information). The CaR-ESI-MS assay was also applied to solutions containing scFv and the isomeric ligands L4 and L5. However, it was found that the ATDs of the deprotonated L4 and L5 ions were nearly identical under the IMS conditions used, with maxima at 7.48 and 7.55 ms (Figure S3 in the Supporting Information). As a result, it was not possible to use IMS (as implemented in the current study) to quantify the relative abundances of the bound ligands. Instead, CID was carried out on the mixture of L4 and L5 ions in the transfer region and was found to produce a number of fragment ions: Y2α (m/z 355.2) and Y2 ions (m/z 323.2), Z2α (m/z 337.2) and Z2 ions (m/z 305.2), B1α (m/z 129.1) and B1 ions (m/z 161.1), as well as C1 (m/z 179.1) and Y1 ions (m/z 193.1) (Figure 3b). Additional

Figure 3. Illustrative CID mass spectra (transfer voltage 16 V) measured for (a) deprotonated L4, (b) deprotonated L4 and L5 ions following their release from (scFv + L)9− ions produced by ESI from a solution containing 5 μM of each of scFv, L4, and L5, and (c) deprotonated L5. Figure 4. ATDs measured for deprotonated L6 and L7, following their release from (a, d, g) the (Pref + L)6− ions, (b, e, h) the (TcdB-B3 + L)11− ions, and (c, f, i) the (TcdB-B3 + L)10− ions (where L = L6 or L7) produced by ESI from solutions of TcdB-B3 (8.2 μM), Pref (15 μM) and (a−c) L6 (4 μM) and L7 (4 μM), (d−f) L6 (4 μM) and L7 (8 μM), and (g−i) L6 (4 μM) and L7 (12 μM). Dashed curves correspond to calculated ATD based on the fractional abundance of L6 and L7 of (a) 0.59 and 0.41, (b) 0.56 and 0.44, (c) 0.73 and 0.27, (d) 0.33 and 0.67, (e) 0.42 and 0.58, (f) 0.47 and 0.53, (g) 0.26 and 0.74, (h) 0.35 and 0.65, and (i) 0.34 and 0.66, respectively.

fragment ions corresponding to water loss were also observed (e.g., m/z 175.0 and 157.1 from the Y1 ion, m/z 161.1 and 143.1 from the C1 ion). CID of the individual L4 and L5 ions produced identical fragment ions, although with different relative abundances (Figure 3a,c). As a result, it was possible to establish the relative abundances of the released L4 and L5 ions from a comparison with CID mass spectra acquired for the individual ions. Applying the approach described in the Experimental Section to the common Z2 fragment ion (m/z 7642

dx.doi.org/10.1021/ac401627t | Anal. Chem. 2013, 85, 7637−7644

Analytical Chemistry

Editors' Highlight

B3 + L)11− (Figure 4b), and (TcdB-B3 + L)10− ions (Figure 4c) at a trap voltage of 50 V. The ATDs clearly exhibit two features, centered at 10.12 and 11.33 ms. These drift times are consistent with the maxima observed in the ATDs measured for the deprotonated ions of L6 (10.12 ms) and L7 (11.33 ms) produced directly from solution (data not shown). The relative SATD,L6 and SATD,L7 peak areas were obtained for the L6 and L7 ions released from the (TcdB-B3 +L) ions at the 10− and 11− charge states (Figure 4b,c). From the average SATD,L6 and SATD,L7 values (determined from both charge states), apparent values f L6* and f L7* of 0.52 ± 0.07 and 0.48 ± 0.07, respectively, were calculated. Similarly, from the average SATD,L6,ns and SATD,L7,ns values, f L6,ns and f L7,ns were found to be 0.57 ± 0.02 and 0.43 ± 0.02, respectively. It follows that the corrected (for nonspecific binding) fractional abundances f L6,sp and f L7,sp are 0.70 ± 0.05 and 0.30 ± 0.02, respectively. From these values, Ka values were calculated to be (7.4 ± 0.9) × 103 M−1 and (3.1 ± 0.3) × 103 M−1 for L6 and L7, respectively. These results agree very well with the affinities obtained using the direct ESI-MS assay. Measurements performed at different ligand concentrations (Figure 4d−4i) yielded similar results (Table 3). Importantly, the Ka values were found to be independent of the trap voltage used (Table 3 and Table S3 in the Supporting Information).

ligand bound to the protein was determined from the relative abundance of fragment ions produced by CID.

■

* Supporting Information Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

8.2 8.2 8.2 a

[L6]o (μM) 4 4 4

[L7]o (μM) 4 8 12 average

Ka,L6 (M−1)

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors acknowledge the Alberta Glycomics Centre and the Canada Foundation for Innovation for financial support and Professors T. Lowary and D. Bundle (University of Alberta) and Professor K. Ng (University of Calgary) for generously providing carbohydrate ligands and proteins used in this study.

■

(7.4 (8.9 (9.1 (8.5

± ± ± ±

0.9) 0.8) 1.0) 1.0)

× × × ×

10 103 103 103

(3.1 (3.4 (6.7 (4.4

± ± ± ±

0.3) 0.4) 0.8) 1.1)

× × × ×

REFERENCES

(1) Williams, S. J.; Davies, G. J. Trends Biotechnol. 2001, 9, 356−362. (2) Fukui, S.; Feizi, T.; Galustian, C.; Lawson, A. M.; Chai, W. Nat. Biotechnol. 2002, 20, 1011−1017. (3) Plante, O. J.; Palmacci, E. R.; Seeberger, P. H. Science 1998, 291, 1523−1527. (4) Plante, O. J.; Seeberger, P. H. Curr. Opin. Drug Discovery Dev. 2003, 6, 521−525. (5) Disney, M. D.; Seeberger, P. H. Drug Discovery Today: Targets 2004, 3, 151−158. (6) Rillahan, C. D.; Paulson, J. C. Annu. Rev. Biochem. 2011, 80, 797− 823. (7) Patwa, T.; Li, C.; Simeone, D. M.; Lubman, D. M. Mass Spectrom. Rev. 2010, 29, 830−844. (8) Lepenies, B.; Seeberger, P. H. Immunopharmacol. Immunotoxicol. 2010, 32, 196−207. (9) Ban, L.; Pettit, N.; Li, L.; Stuparu, A. D.; Cai, L.; Chen, W.; Guan, W.; Han, W.; Wang, P. G.; Mrksich, M. Nat. Chem. Biol. 2012, 8, 769− 773. (10) Rogersa, C. J.; Clarka, P. M.; Tullya, S. E.; Abrol, R.; Garcia, K. C.; Goddard, W. A., III; Hsieh-Wilson, L. C. Proc. Natl. Acad. Sci. U.S.A. 2011, 108, 9747−9752. (11) Nimrichter, L.; Gargir, A.; Gortler, M.; Altstock, R. T.; Shtevi, A.; Weisshaus, O.; Fire, E.; Dotan, N.; Schnaar, R. L. Glycobiology 2004, 14, 197−203. (12) Smith, D. F.; Song, X.; Cummings, R. D. Methods Enzymol. 2010, 480, 417−444. (13) El-Hawiet, A.; Shoemaker, G.; Daneshfar, R.; Kitova, E. N.; Klassen, J. S. Anal. Chem. 2012, 84, 50−58. (14) Cederkvist, F. H.; Zamfir, A D.; Bahrke, S.; Eijsink, V.G. H.; Sørlie, M.; Peter-Katalinicì, J.; Peter, M. G. Angew. Chem., Int. Ed. 2006, 45, 2429−2434. (15) Abzalimov, R. R.; Dubin, P. L.; Kaltashov, I. A. Anal. Chem. 2007, 79, 6055−6063. (16) Williams, J. P.; Grabenauer, M.; Holland, R. J.; Carpenter, C. J.; Wormald, M. R.; Giles, K.; Harvey, D. J.; Bateman, R. H.; Scrivens, J. H.; Michael T. Bowers, M. T. Int. J. Mass Spectrom. 2010, 298, 119− 127. (17) Fenn, L. S.; McLean, J. A. Phys. Chem. Chem. Phys. 2011, 13, 2196−2205. (18) Clowers, B. H.; Dwivedi, P.; Steiner, W. E.; Hill, H. H., Jr.; Bendiak, B. J. Am. Soc. Mass Spectrom. 2005, 16, 660−669. (19) Li, H.; Giles, K.; Bendiak, B.; Kaplan, K.; Siems, W. F.; Hill, H. H., Jr. Anal. Chem. 2012, 84, 3231−3239.

Ka,L7 (M−1) 3

AUTHOR INFORMATION

Corresponding Author

Table 3. Association Constants (Ka) for TcdB-B3 Binding to L6 (Ka,L6) and L7 (Ka,L7) Measured in Aqueous Ammonium Acetate (pH 7 and 24 °C) at Different Initial Ligand Concentrations Using the CaR-ESI-MS Assaya [TcdBB3]o (μM)

ASSOCIATED CONTENT

S

103 103 103 103

Error corresponds to one standard deviation.

■

CONCLUSIONS The CaR-ESI-MS assay has emerged as a powerful tool for screening carbohydrate libraries against soluble proteins to identify specific interactions. In contrast to glycan array-based screening, the CaR-ESI-MS approach allows for both the detection and quantification of ligands present in mixtures of hundreds of carbohydrates. The present results demonstrate the applicability of the CaR-ESI-MS assay to distinguish and quantify isomeric carbohydrate ligands. Absolute affinities for isomeric ligands were determined from the abundance ratio of ligand-bound to free protein measured directly by ESI-MS and the relative abundances of the bound ligands, which were established by releasing the ligands, in their deprotonated form, from the protein using collision-induced dissociation (CID) and subjecting them to ion mobility separation (IMS) or another stage of CID. Using Gaussian functions to represent the contribution of each ligand to the arrival time distributions (ATDs) measured by IMS, the relative abundance of each ligand bound to the protein was established. A modified form of this method, suitable for cases where nonspecific ligand-protein binding occurs during the ESI process, was also developed. In cases where the ATDs are not sufficiently different to distinguish isomeric ligands, the relative abundance of each 7643

dx.doi.org/10.1021/ac401627t | Anal. Chem. 2013, 85, 7637−7644

Analytical Chemistry

Editors' Highlight

(20) Winkler, W.; Huber, W.; Vlasak, R.; Allmaier, G. Rapid Commun. Mass Spectrom. 2011, 25, 3235−3244. (21) Yamaguchi, Y.; Nishima, W.; Re, S.; Sugita, Y. Rapid Commun. Mass Spectrom. 2012, 26, 2877−2884. (22) Yamagaki, T.; Sato, A. Anal. Sci. 2009, 8, 985−988. (23) Li, H.; Bendiak, B.; Siems, W. F.; Gang, D. R.; Hill, H. H., Jr. Anal. Chem. 2013, 85, 2760−2769. (24) Zekavat, B.; Solouki, T. J. Am. Soc. Mass Spectrom. 2012, 23, 1873−1884. (25) Rademacher, C.; Shoemaker, G. K.; Kim, H. S.; Zheng, R. B.; Taha, H.; Liu, C.; Nacario, R. C.; Schriemer, D. C.; Klassen, J. S.; Peters, T.; Lowary, T. L. J. Am. Chem. Soc. 2007, 129, 10489−10502. (26) Zdanov, A.; Li, Y.; Bundle, D. R.; Deng, S.-J.; MacKenzie, C. R.; Narang, S. A.; Young, N. M.; Cygler, M. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 6423−6427. (27) Wang, W.; Kitova, E. N.; Klassen, J. S. Anal. Chem. 2005, 77, 3060−3071. (28) Sun, J.; Kitova, E. N.; Wang, W.; Klassen, J. S. Anal. Chem. 2006, 78, 3010−3018. (29) Deng, L.; Kitova, E. N.; Klassen, J. S. J. Am. Soc. Mass Spectrom. 2013, 24, 988−996.

7644

dx.doi.org/10.1021/ac401627t | Anal. Chem. 2013, 85, 7637−7644