Bioconjugate Chem. 1994, 5, 194- 198
194
Characterization of Hapten Binding to Immunoconjugates by Electrospray Ionization Mass Spectrometry Kenneth M. Straub’J and Mark J. Levy’ Syntex Discovery Research and Syva Co., 3401 Hillview Avenue, Palo Alto, California 94304. Received November 18, 1993”
Electrospray ionization mass spectrometry (ESI-MS) has been used to characterize the covalent binding of different haptens to the enzyme glucose 6-phosphate dehydrogenase. The technique allows one to directly observe the relative amounts of each conjugated species present in a mixture, as well as the average hapten number. These measurements are useful for optimizing reaction conditions to yield a more precisely defined product for use in immunoassays. The results obtained show that ESI-MS with a quadrupole analyzer can be successfully used to analyze mixtures of derivatized proteins in the molecular weight range of 50-60 kDa, where the modifying hapten has a molecular weight as low as 130 Da.
INTRODUCTION There are numerous examples in protein chemistry where it is desirable to covalently modify the side chains of amino acid residues to give specifically derivatized macromolecules. For example, immunoconjugates comprised of haptens linked to carrier proteins are useful for eliciting the production of specific antibodies for use in immunoassays, and hapten-modified enzymes form the basis of many types of enzyme immunoassays. Conjugation of proteins with low molecular weight haptens gives a mixture of macromolecules that differ in the number and site of modified residues. The hapten number n describes the average number of covalently modified residues per protein and is a key parameter that must be carefully optimized when preparing enzyme tracers for use in immunoassays. A number of analytical methods have been described that allow the determination of an average hapten number. Classical methods include the use of absorption spectroscopy and radiolabeled hapten. These techniques, while simple to apply, give an average hapten number but do not provide information on the actual distribution of modified macromolecules. These techniques can also overestimate the hapten number if noncovalently bound hapten is present in the sample. In addition, these methods suffer from a number of other limitations. Absorption spectroscopy requires that the hapten have a significant extinction coefficient over a useful range of wavelengths and is subject to uncertainty about changes in the absorbance of bound chromophores. The use of radiolabeled haptens requires the preparation of labeled material with a high specific activity. Recently, matrix-assisted laser desorption ionization (MALDI) mass spectrometry has been used to estimate the hapten number of covalently modified proteins (1,2). While this methodology appears to be a substantial improvement over conventional assays, the low mass resolution of the technique limits it to the determination of an average hapten modification number.
* To whom correspondence should be addressed. Tel (415) 855-5067;FAX (415)354-7363. + Syntex Discovery Research. Syva Co. e Abstract published in Advance ACS Abstracts, March 15, 1994.
*
In principle, electrospray mass spectrometer (ESI-MS) is an ideal technique for this type of problem (3). Since ESI-MS can be implemented on quadrupole or magnetic sector mass spectrometers, the increased resolution afforded by these systems allows one to directly observe the relative amounts of each conjugated species present in a mixture, as well as the average hapten number. Knowing the amounts of each species present can help optimize reaction conditions to yield a more precisly defined product for use in immunoassays. We have explored the use of ESI-MS to characterize the stoichiometry of covalent binding of different haptens to the enzyme glucose 6-phosphate dehydrogenase (G6PDH, EC 1.1.1.49). The homodimeric subunit of this enzyme has a molecular weight of 54.5 kDa and is widely used in enzyme-linked immunoassays. The results obtained show that ESI-MS with a quadrupole analyzer can be successfully used to analyze mixtures of derivatized proteins in the molecular weight range of 50-60 kDa where the modifying hapten has a molecular weight as low as 130 Da. EXPERIMENTAL SECTION Materials and Methods. G6PDH was purchased from US Biochemicals. Digoxigenin 3-0-hemisuccinate succinimide ester (Figure 1, structure 2) was obtained from Boehringer-MannheimBiochemicals. Desaminothyroxine (structure I), 2,2-dipropylgutaric anhydride (structure 3), and 5-[N-(2’-carboxyethyl)carbamoylldibenz[b,flazepine (structure 4) were synthesized as previously reported (46 ) . Tritiated 5-[N-(2’-carboxyethyl)carbamoylldibenz[b,flazepinehad a specific activityof 7 Ci/mmol. All other chemicals were reagent grade or better. Preparation of Immunoconjugates. The acidic haptens were activated for conjugation by formation of N-succinimidyl esters (7). Conjugation reactions were conducted in solutions containing 5 mg/mL of GGPDH, 100 mM NaHC03 (pH 8.0), and 20% dimethylformamide (DMF). Activated hapten dissolved in DMF was added to the enzyme solution while the solution was stirred on ice. A 25-fold molar excess (relative to G6PDH subunit) of desaminothyroxine and 10-fold excesses of digoxigenin and 2,2-dipropylglutaric anhydride were used in the conjugations. For the 5-[N-(2’-carboxyethyl)carbamoylldibenz[b,flazepine conjugations, the molar excess was varied from 5-fold to 20-foldto achieve avariety of loadings.
1043-1802/94/2905-0194~04.50~~ 0 1994 American Chemical Society
Characterization of Hapten Binding by ESI-MS
Unreacted hapten was removed after 1h by chromatography over Biogel P6-DG resin (Bio-Rad) in 10 mM ammonium acetate buffer (pH 6.7). Residual salts were removed from the samples of three cycles of (1)dilution to 2 mL with 10 mM ammonium acetate buffer and (2) concentration to approximately 100 p L using Centricon30 microconcentrators (Amicon). Labeling density of radioactive conjugates was determined by conventional methods. Aliquots of the purified conjugates were diluted to approximately 10 pg/mL in water. Duplicate 800-pL samples were taken from these common solutions for use in scintillation counting and protein assay. Protein assays were performed according to the method of Bradford by adding 200 pL of dye reagent (Bio-Rad) to the samples and measuring absorbance at 595 nm (8). Unmodified G6PDH was used to generate a standard curve. Radioactivity was measured by adding 10 mL of Ready-Safe (Beckman) to each sample and counting on a Beckman scintillation counter. Mass Spectrometry. Electrospray ionization mass spectra were obtained using a Finnigan-MAT TSQ700 (Finnigan-MAT) equipped with an electrospray ion source (Analytica of Branford). Samples were prepared at concentrations of 1-5 pmol/pL in 10 mM NH40Ac containing 20-50 % acetonitrile or methanol and infused directly into the source at 1-2 pL/min. Ion source operating conditions were as follows: cylindrical electrode, -3.0 kV; capillary voltage, +50 V; tube lens, 150-195 V; drying gas temperature, 130 OC; drying gas flow, ca. 6 L/min. The instrument was scanned over the m/z range 300-2000 in 3-9 intervals. All data were acquired in profile mode, and successive scans were averaged until the required signal-to-noise ratio was obtained. Typically, 5-10-min averaging times were required to obtain useful data on samples at concentrations of 1-5 pmol/pL and an infusion rate of 1pL/min, corresponding to 5-50 pmol of sample consumed for each measurement. Alternatively, samples in 0.05% trifluoroacetic acidwater were electrosprayed using a sheathing flow of 2-methoxyethanol (1 pL/min) and a nebulizing flow of nitrogen (600mL/min). Operating conditions were similar to what was employed for methanol-water or acetonitrilewater solutions, except that a higher drying gas temperature (255 “C) and drying gas flow rate (40 mL/min) were employed. These conditions are useful for analyzing modified proteins that have reduced solubility in methanol-water or acetonitrile-water, Multiply charged spectra were transformed to give molecular weights using the deconvolution program based on the algorithm described by Mann et a1 (9). An implementation of this algorithm is supplied with the “BioSoft” software on the TSQ700 data system. The hapten number (n)was calculated from the observed molecular weight distribution as the sum of the product i and its hapten of the intensities I of each species j number n divided by the sum of the intensities for all components:
-
1
C J=on j I j n=i
CIj J’o
RESULTS
The general utility of ESI-MS for characterizing covalently modified proteins is demonstrated by the results of experiments using the enzyme glucose 6-phosphate
Bioconjugate Chem., Voi. 5, No. 3, 1994
105
R-2-o-a+ Desaminothyroxine hapten 1
Am = 744 Da
0
Digoxigenin hapten 2
Am = 472 Da
Valproic Acid hapten 3
Am = 198 Da
Carbamazepine hapten 4
Am = 290 Da
Figure 1. Structuresof haptensused for conjugationwith glucose 6 phosphate dehydrogenase. Residue masses for each hapten following conjugation reaction are indicated.
dehydrogenase (GGPDH). This enzyme is widely used in commercially available enzyme-linked immunoassays. This type of assay requires covalently binding hapten to an enzyme while minimizing loss of catalytic activity. A number of different G6PDH conjugateswere prepared by using carboxylicacid-containing haptens with molecular weights of 300-700. The structures of these haptens are shown in Figure 1 and include analogues of thyroxin (structure 11, digoxigenin (structure 21, valproic acid (structure 31, and carbamazepine (structure 4). Covalent binding of these haptens to GGPDH, primarily by reaction with t-amino groups of lysine residues, yields a series of mixtures of modified proteins. Each of these proteins differs in molecular weight by the residue mass of the hapten. The corresponding residue mass differences for these haptens are 744 Da (thyroxin analogue), 472 Da (digoxigenin analogue), 198 Da (valproate analogue), and 290 Da (carbamezepine analogue). When electrosprayed from a 30% methanol solution, GGPDH gives a typical ESI mass spectrum with a distribution of multiply charged species between +28 and
196
Straub and Levy
BioconJugate Chem., Vol. 5, No. 3, 1994 I I I I I I II II II II
1007
144’
II 1 1 ~ ~ 5 1 1I1
II
II
11
-
II I
1202 1273 j 13[:454 918
80
80
p
8
.*
f
60
f
ga
$
60
2 a 40
40
20
20
800
1000
1200
1400
1600
1800
600
800
1000
1400
1200
1600
I
I
1‘ 55224
1001
45000 45000
50000
55000
1800
d z
d z
60000
mw
Figure 2. (A) ESImass spectrum of GGPDH. (B) Deconvoluted spectrum obtained from data in A.
50000
55000
r
60000
65000
mw
Figure 3. (A) ESI mass spectrum of GGPDH following conjugation with thyroxine (analogue l). (B)Deconvoluted spectrum obtained from data in A. Hapten number of each component is
indicated.
+70, as shown in Figure 2A. The multiply charged species are well resolved in this spectrum, and transformation of the data to molecular weight can be carried out by inspection or by application of a suitable algorithm such as that described by Mann et al. (9). Application of this algorithm gives the data presented in Figure 2B, which show that a single component of MW 54477 f 5 Da is present. Modificationof G6PDH by the thyroxin analogue 1gives a mixture of proteins that yields an ESI mass spectrum in which each of the modified proteins independently contributes a series of multiply charged ions. The complexity of the resulting spectrum depends on both the modification number and the relative mass of the hapten. In the case of a low modificationnumber and a large residue mass, the multiply charged ion series can still be resolved, and the deconvoluted spectrum shows a simple product distribution. For the thyroxin analogue, a mixture consisting predominantly of Thyro, Thyrl, and Thyrz is obtained, where n = 1.2 (Figure 3). As shown in Figure 3A, the raw data for the thyroxin conjugate consist of three to four sets of superimposed multiply charged species. A molecular weight profile can be determined either by inspection or by application of a transformation function. As the mass of the hapten species decreases, the multiply charged species begin to overlap, eventually reaching a point where individual components can no longer be resolved. Simple inspection of the data is no longer sufficient for assigning multiply charged ions to the different series. This is shown by the data in Figures 4 and 5. For the digoxigenin-labeled enzyme (residue mass 472 Da; Figure 4), a series of products is obtained where n = 3-8. These products have an average modification
number of 6.4. The mass difference between the components corresponds to 472 f 4 Da, as expected for this hapten. In the case of the valproate-labeled enzyme (Figure 51, the low residue mass (Am = 198) and increased degree of labeling results in significant overlap between the multiply charged species. Despite this overlap, complete deconvolution of the electrospray data to yield a molecularweight distribution is still possible, as shown in Figure 5B. In this example, components corresponding to n = 0-5 haptens are present, with an average modification number of 2.1. Note that the mass differences between the adjacent peaks are within the expected error of 0.01 76. Heterogeneity in the enzyme or protein can cause additional complexity. A common source of such heterogeneity is the partial removal of a methionine residue from the amino terminus of proteins that have been produced by recombinant techniques in bacteria, so that a mixture of two species differing by 131 Da is present. Figure 6, obtained from a digoxigenin analog-labeled G6PDH present in both Met and des-Met forms, is an example. The data shows that there are 10species present, corresponding to five pairs of digoxigenin-labeledG6PDH where n = 1-5 with and without an N-terminal methionine. Comparison of Hapten Numbers Determined by ESI-MS and Radiolabeling. Labeling density for G6PDH-3H-5-[N-(2’-carboxyethyl)carbamoylldibenz b,flazpeine conjugate was determined for three different loadings of the hapten both by ESI-MS and by scintillation counting. Measurement of radioactivity and protein concentration yielded 1.5,2.9, and 4.2 moieties per subunit for the three different conjugates A, B, and C. The
Characterization of Hapten Binding by ESI-MS
Bloconjugate Chem., Vol. 5, No. 3, 1994 I
I
'"1
1000
800
1200
1400
1600
1
1800
1 A
600
m/z
800
1000
1200
I
t3 s
1
80
-
1400
1600
1800
d Z
100,
I
loo
197
80 Am = 472
1
B
1
I
54874 1
1(
r
54675
Am
= 798
r g.5072
60-
2
d
4020
I
,
,
45000
,
,
I
50000
55000 mw
60000
65000
50000
52000
54000
56000
58000
60000
mw
Figure 4. (A) ESI mass spectrum of GGPDH following conjugation with digoxigenin (analogue 2). (B)Deconvoluted spectrum obtained from data in A. Hapten number of each component is indicated.
Figure 5. (A) ESI mass spectrum of GGPDH following conjugation withvalproic acid (analogue3). (B)Deconvoluted spectrum obtained from data in A. Hapten number of each component is indicated.
respective labeling densities derived from the mass spectrometric determinations were 1.2,2.6, and 3.4 (Table 1).
In many instances, the major factor limiting resolution is not the analyzer performance but the presence of noncovalently bound adducts that contribute to broadening and attendant overlap of the multiply charged sample peaks. Alkali cations, primarily sodium, are particularly troublesome is this respect. Removal of these interfering substances using purification techniques such as ultrafiltration is sometimes successful, but severe loss of sample by irreversible adsorption limits this approach with many conjugated proteins. The discrepancy observed between the hapten number as determined by the ESI-MS and radiolabeling methodologies can be attributed to the presence of noncovalently bound hapten in the samples despite extensive sample purification. This discrepancy would result in an overestimation of conjugation by the radiolabeling technique. -A similar overestimation of hapten number as determined by absorption spectroscopy was noted by Wengatz et al. in their description of the use of MALDI for characterizing modified proteins (I). While it is also possible that ionization efficiencies may differ for proteins with different modification numbers, the evidence to date suggests that this is not the case. For example, it was observed during the course of these studies that the overall response factors for G6PDH modified with haptens 1-4 were essentially equivalent, despite the diversity of these functional groups. Another possible source of error is that proteins that are heavily modified with nonpolar moieties could have reduced solubility in the solvent used for electrospray ionization, so that components with large values of n would be lost prior to analysis. For the carbamazepine hapten, loadings that resulted in hapten
DISCUSSION
With a quadrupolar analyzer, a practical limit to analyzing these types of mixtures appears to occur for a 55-kDa protein at a modification number of ca. 8 and a hapten mass of approximately 130 Da. Mixtures that are more complex than this result in a degree of overlap between the various multiply charged species such that transformation of the data to yield a molecular weight profile becomes impractical. Improved mass resolution is the most obvious way to extend these measurements to more complex mixtures. Coupling ESI to magnetic sector or ICR-type instruments should result in improved resolution of the multiply charged ions contributing to each series (10, 11). For a quadrupole analyzer, increasing mass resolution does not always yield substantial improvements in the deconvoluted data since the resulting decrease in sensitivity offsets any real gain in signal. It is sometimes possible to improve the quality of the data by employing conditions that result in a shift of the entire envelope of multiply charged ions to higher mlz (lower charge number), where the mlz difference between successive charged species is larger. For example, use of water containing 0.1 5% trifluoroacetic acid in combination with 2-propanol or 2-methoxyethanol as a sheathing liquid will shift the distribution of charged ions for GGPDH from +40 out to +30,with ions observed up to mlz 3500. This technique is limited by the decreased sensitivity and resolution of quadrupole systems above mlz 2000.
Straub and Levy
Bioconjugate Chem., Vol. 5, No. 3, 1994
198
loo
1
1568
A
80
3 3
5
s ? 2
8
60
40
20
800
1000
1200
1400
1600
1800
m/z
I
1
The MALDI technique appears to be promising for determining the average modification number of conjugates, although the low mass resolution inherent to this type of analyzer makes resolution of individual conjugated species impractical. While some information about the distribution of species in a mixture can be obtained from a consideration of the peak width of the resulting data, such measurements can be misleading since noncovalently bound gas-phase clusters derived from sample matrix or other contaminants can contribute to increasedpeak width. It is unlikely, for example, that the presence of a des-Met analogue as shown in Figure 6 could have been detected by the MALDI technique. In summary, ESI-MS is a relatively simple but powerful technique for characterizing covalently modified protein conjugates. Information can be readily obtained that describes both the average modification number as well as the relative amount of each species present in the sample. ACKNOWLEDGMENT The authors thank Dr. Michael Huster for the synthesis of desaminothyroxineand tritiated 5- [N-(2’-carboxyethyl)carbamoylldibenz[b,flazepineand Dr. Thomas Goodman for a sample of recombinant G6PDH. LITERATURE CITED (1) Wengatz,I., Schmid,R. D., Kreissig,S., Wittmann, C., Hock, B., Ingendoh, A., and Hillenkamp, F. (1992) Determination of the hapten density of immunoconjugates by matrix-assisted UVlaser desorption/ionizationmass spectrometry. Anul.Lett.
52000
54000
56000
58000
60000
mw
Figure 6. (A) ESI mass spectrum of a mixed G6PDH with and without an N-terminal methionine followingconjugation with a digoxigenin analogue. (B) Deconvoluted ESI mass spectrum obtained from data in A. The hapten number of each component is indicated; desMet-forms are indicated with asterisk. Table 1. Comparison of Hapten Numbers Determined by ESI-MSand Radiolabeling for Carbamazepine-Modified GGPDH
method scintillation counting
samplea
ESI-MS
A
1.2
1.5
B
2.6
2.9
4.2 C 3.5 @Sampleswith different hapten numbers were prepared as described in the Experimental Section. numbers greater than n = 6 resulted in a product mixture that precipitated in the presence of 30% acetonitrile or methanol. This can be guarded against by monitoring the absorbance of the solution after centrifugation or ultrafiltration to verify that the total protein concentration has not changed. Comparing the quality of the information obtained by ESI-MS with that afforded by conventional methods for determining hapten number is instructive. Absorption spectroscopy is restricted to haptens that absorb in a useable range of wavelengths, and the results are subject to uncertainty regarding changes in the extinction coefficient that bound material could undergo. As noted above, both absorption spectroscopy and the use of radiolabeled hapten can be complicated by the presence of noncovalently bound hapten that is not removed by the usual size-exclusion techniques. While it is possible to employ conditions in electrospray ionization that will preserve noncovalent interactions, the methodology described here yields ions from covalently bound species only (12).
25,1993-1997. (2) Siegel, M. M., Hollander, I. J., Hamann, P. R., James, J. P.,
Hinman, L., Smith, B. J., Farnsworth, A. P. H., Phipps, A., King,D. J., Karas, M., Ingendoh,A,,and Hillenkamp,F. (1991) Matrix-assisted UV-laser desorption/ionization mass spectrometric analysis of monoclonalantibodies for the determination of carbohydrate, conjugated chelator, and conjugated drug content. Anal. Chem. 63, 2470-2481. (3) Smith, R. D., Loo,J. A., Edmonds,C. G., Barinaga, C. J., and Udseth, H. R. (1990) New developments in biochemicalmass spectrometry: electrospray ionization. Anal. Chem. 62,882899. (4) Kharasch, N., Kalfayan, S. H., and Arterberry, J. D. (1956)
The synthesis of some methyl analogs of desaminothyroxine. J. Org. Chem. 21, 925-930. (5) Leung, D. K., and Singh, P. (1980) Valproate conjugation using dicarbonyls. US. Patent 4,238,389. (6) Singh, P. (1977) Tegretol antigens and antibodies. U.S. Patent 4,058,511. (7) Anderson, G. W., Zimmerman, J. E., and Callahan, E. M. (1964) The use of esters of N-hydroxysuccinimidein peptide synthesis. J. Am. Chem. SOC. 86, 1839. (8) Bradford, M. M. (1976) A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principles of protein-dye binding. Anal. Biochem. 72, 248254. (9) Mann, M., Meng, C. K., and Fenn, J. B. (1989) Interpreting mass spectra of multiply charged ions. Anal. Chem. 61,17021708. (10) Meng, C.-K., McEwen, C. N., and Larsen, B. (1990)
Electrosprayionizationon a high-performancemagneticsector mass spectrometer. Rapid Commun. Mass Spectrom. 1,147150. (11) Beu, S. C., Senko, M. W., Quinn, J. P., Wampler, F. M., and McLafferty, F. W. (1993) Fourier-transform electrospray
instrumentation for tandem high resolutionmass spectrometry of large molecules. Rapid Commun. Mass Spectrom. 4,557-
565. (12) Li, Y.-T., Hsieh, Y.-L., Henion, J. D., and Ganem, B. (1993)
Studies on heme binding in myoglobin, hemoglobin, and cytochrome c by ion spray mass spectrometry. J. Am. SOC. Mass Spectrom. 4, 631-637.
