Calicheamicin Derivatives Conjugated to Monoclonal Antibodies

Wyeth-Ayerst Research, Lederle Laboratories, Pearl River, New York 10965. Stefan Berkenkamp and Franz Hillenkamp. Institute of Medical Physics, Univer...
0 downloads 0 Views 493KB Size
Download PDF
Anal. Chem. 1997, 69, 2716-2726

Calicheamicin Derivatives Conjugated to Monoclonal Antibodies: Determination of Loading Values and Distributions by Infrared and UV Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry and Electrospray Ionization Mass Spectrometry Marshall M. Siegel,* Keiko Tabei, Arthur Kunz, Irwin J. Hollander, Philip R. Hamann, and Duncan H. Bell

Wyeth-Ayerst Research, Lederle Laboratories, Pearl River, New York 10965 Stefan Berkenkamp and Franz Hillenkamp

Institute of Medical Physics, University of Mu¨ nster, D-4400 Mu¨ nster, Germany Calicheamicin derivatives (MW ∼1500) and monoclonal antibodies (MoAbs) conjugated to calicheamicin derivatives (MW ∼150 000) were analyzed by UV-MALDI/MS, IR-MALDI/MS, and ESI/MS. These materials are potent anticancer agents. Calicheamicin derivatives and conjugates rapidly degrade upon UV irradiation but are relatively stable during IR irradiation and under ESI conditions. A unique feature of IR-MALDI/MS is a 2 times enhancement in resolution relative to UV-MALDI/MS for masses above ∼50 000 Da resulting in a molecular ion envelope containing a series of partially resolved peaks of the calicheamicin-MoAb conjugates. The mass shift difference between the peak maxima corresponded to the mass change due to the covalent addition of calicheamicin derivatives to the monoclonal antibody. The distribution of the calicheamicin derivatives in the monoclonal antibodies was computed by deconvoluting the partially resolved peak envelope. A unique feature of the ESI mass spectra, under unit resolution conditions, is that the distribution of the carbohydrates can be well resolved for pure MoAbs and can be only partially resolved for conjugated MoAbs. Average loading values for calicheamicin derivatives when conjugated to MoAbs were computed from UV-MALDI/MS, IR-MALDI/MS, and ESI/ MS data and the results compared with the average loading values obtained by UV absorption spectrometry. Very low average loading values were computed from UVMALDI/MS data due to the degradation of the conjugated calicheamicin derivatives during the UV irradiation process. The IR-MALDI/MS average loading values, obtained with glycerol as the matrix, were consistent with the UV absorption spectrometry values for conjugates having hydrolytically stable linkers, but not when the linker contained a hydrolytically labile hydrazone. ESI/ MS average loading values were generally lower than the corresponding values obtained by IR-MALDI/MS. The average loading values and distributions obtained using

IR-MALDI/MS were more reliable than the corresponding ESI/MS values because the partially resolved, singly and doubly charged peaks in the IR-MALDI spectra can be mathematically deconvoluted, while the overlapping, highly multiply charged peaks of the electrospray spectra can only be partially deconvoluted.

2716 Analytical Chemistry, Vol. 69, No. 14, July 15, 1997

S0003-2700(97)00035-8 CCC: $14.00

Low molecular weight drugs covalently attached to immunogenic proteins are used to elicit antibody responses in animals1 and, when covalently attached to appropriate antibodies, are being evaluated as a means for delivering anticancer drugs to tumor sites in humans.2 The average number of moles of drug relative to the number of moles of protein or antibody carrier, referred to as the average loading value, is usually determined by UV absorption spectrometry, radioactivity, immunoassay, and/ or wet chemical measurements. None of these methods are capable of determining the distribution of the numbers of drug molecules covalently bound to the carrier molecules. To confirm the results of these techniques and to supplement them when they are inappropriate or not feasible, mass spectrometric techniques are being developed to determine both the average loading values as well as the distributions of drugs conjugated to proteins and antibodies. Matrix-assisted laser desorption/ionization mass spectrometry (MALDI/MS)3-15 and electrospray ionization mass * Address correspondence to this author: 401 N. Middletown Rd., Bldg. 222, Room 1043, Wyeth-Ayerst Research, Lederle Laboratories, Pearl River, NY 10965. Telephone: (914) 732-3417. FAX: (914) 732-3659. E-mail: siegelm@ war.wyeth.com. (1) Collier, R. J.; Kaplan, D. A. Sci. Am. 1984, 251, 56-64. (2) Koppel, G. A. Bioconjugate Chem. 1990, 1, 13-23. (3) Siegel, M. M.; Hollander, I. J.; Hamann, P. H.; James, J. P.; Hinman, L.; Smith, B. J.; Farnsworth, A. P. H.; Phipps, A.; King, D. J.; Karas, M.; Ingendoh, A.; Hillenkamp, F. Anal. Chem. 1991, 63, 2470-2481. (4) Wengatz, I.; Schmid, R. D.; Kreissig, S.; Wittmann, C.; Hock, B.; Ingendoh, A.; Hillenkamp, F. Anal. Lett. 1992, 25, 1983-1997. (5) Siegel, M. M.; Tsou, H. R.; Lin, B.; Hollander, I. J.; Wissner, A.; Karas, M.; Ingendoh, A.; Hillenkamp, F. Biol. Mass Spectrom. 1993, 22, 369-376. (6) Shoyama, Y.; Sakata, R.; Murakami, H. Org. Mass Spectrom. 1993, 28, 987988. © 1997 American Chemical Society

spectrometry (ESI/MS)16-21 have been evaluated in a number of laboratories and found in many cases to be routine, reliable, and general methods for determination of the average loading values and distributions for a number of drugs and pharmaceuticals conjugated to protein carriers. A review of these MALDI/MS and ESI/MS results has recently been published.22 The calicheamicins are potent enediyne-containing antitumor antibiotics23 which function by cleaving DNA.24 The methyl trisulfide of these compounds, e.g., NAc-calicheamicin (1, MW 1409), undergoes bioreduction, and the resultant diradical abstracts hydrogen atoms from opposite strands of the DNA. After cleavage of the disulfide bond and cyclization of the resultant thiol via a Michael addition reaction, the end product of this series of reactions for NAc-calicheamicin is NAc-calicheamicin  (2, MW 1333). NAc-Calicheamicin derivatives conjugated to monoclonal antibodies (MoAbs) are being developed as targeting agents for anticancer therapy and have shown promising results in eradicating xenograft tumors in laboratory animals. They are presently undergoing clinical trials in both leukemia and ovarian cancer patients. The normal loading for these clinically used conjugates ranges from 2 to 3 mol of conjugated calicheamicin drug per mole of monoclonal antibody. The average loading values for calicheamicin conjugates measured to date by UV-MALDI/MS were one-half to one-fifth the values obtained by UV absorption spectrometry,3 by immunoassay, and from biological assays.25-27 In addition, the distributions of calicheamicin derivatives conjugated to MoAbs were not obtainable from the UV-MALDI/MS (7) Shoyama, Y.; Fukada, T.; Tanaka, T.; Kusai, A.; Nojima, K. Biol. Pharm. Bull. 1993, 16, 1051-1053. (8) Adamczyk, M.; Buko, A.; Chen, Y.-Y.; Fishpaugh, J. R.; Gebler, J. C.; Johnson, D. D. Bioconjugate Chem. 1994, 5, 631-635. (9) Alexander, A. J.; Dodsworth, D. W.; Kirkley, D. H.; Root, B.; Tous, G. I. Proceedings of the 40th ASMS Conference on Mass Spectrometry and Allied Topics, Washington, DC, May 31-June 5, 1995; pp 941-942. (10) Siegel, M. M.; Kunz, A.; Bell, D.; Berkenkamp, S.; Karas, M.; Ingendoh, A.; Hillenkamp, F. Proceedings of the 42nd ASMS Conference on Mass Spectrometry and Allied Topics, Chicago, IL, May 29-June 3, 1994; p 967. (11) Watson, E.; Shah, B.; DePrince, R.; Hendren, R. W.; Nelson, R. BioTechniques 1994, 16, 278-281. (12) Jaeschke, A.; Fuerste, J. P.; Nordhoff, E.; Hillenkamp, F.; Cech, D.; Erdmann, V. A. Nucleic Acids Res. 1994, 22, 4810-4817. (13) Lapolla, A.; Gerhardinger, C.; Baldo, L.; Fedele, D.; Keane, A.; Seraglia, R.; Catinella, S.; Traldi, P. Biochim. Biophys. Acta 1993, 1225, 33-38. (14) Lapolla, A.; Fedele, D.; Seraglia, R.; Catinella, S.; Traldi, P. Rapid Commun. Mass Spectrom. 1994, 8, 645-652. (15) Goto, Y.; Shima, Y.; Morimoto, S.; Shoyama, Y.; Murakami, H.; Kusai, A.; Nojima, K. Org. Mass Spectrom. 1994, 29, 668-671. (16) Straub, K. M.; Levy, M. J. Bioconjugate Chem. 1994, 5, 194-198. (17) Le Blanc, J. C. Y.; Siu, K. W. M. Anal. Chem. 1994, 66, 3289-3296. (18) Springer, D. L.; Bull, R. J.; Goheen, S. C.; Sylvester, D. M.; Edmonds, C. G. J. Toxicol. Environ. Health 1993, 40, 161-176. (19) Stevens, R. D.; Bonaventura, J.; Bonaventura, C.; Fennel, T. R.; Millington, D. S. Biochem. Soc. Trans. 1994, 22, 543-547. (20) Bolton, J. L.; Le Blanc, J. C. Y.; Siu, K. W. M. Biol. Mass Spectrom. 1993, 22, 666-669. (21) Bruenner, B. A.; Jones, A. D.; German, J. B. Rapid Commun. Mass Spectrom. 1994, 8, 509-512. (22) Siegel, M. M. In Protein and Peptide Analysis by Mass Spectrometry; Chapman, J. R., Ed.; Methods in Molecular Biology 61; Humana Press: Totowa, NJ, 1996; Chapter 15, pp 211-226. (23) Lee, M. D.; Durr, F. E.; Hinman, L. M.; Hamann, P. R.; Ellestad, G. A. Adv. Med. Chem. 1993, 2, 31-66. (24) Ellestad, G. A.; Zein, N.; Ding, W. Adv. DNA Sequence Specific Agents 1992, 1, 293-318. (25) Hinman, L. M.; Wallace, R. E.; Hamann, P. R.; Durr, F. E.; Upeslacis, J. Antibody Immunoconjugates, Radiopharm. 1990, 3, 59 (Abstr. 84). (26) Wallace, R. E.; Hinman, L. M.; Hamann, P. R.; Lee, M.; Upeslacis, J.; Durr, F. E. J. Cancer Res. Clin. Oncol. 1990, 116, 323 (Suppl. Part I). (27) Hamann, P. R.; Hollander, I. J.; Holcomb, R.; Hallett, W.; Beyer, C. F.; Lindh, D.; Hinman, L. M. Antibody Immunoconjugates, Radiopharm. 1995, 8, 66 (Abstr. 29).

data. These poor correlations with accepted assays prompted us to seek out the causes of these failures. The following studies demonstrate the lability of calicheamicin derivatives and calicheamicin-MoAb conjugates under UV laser irradiation and show that more reliable calicheamicin loading values as well as drug loading distributions can be obtained using IR lasers in the MALDI/MS experiments. Also, the limitations of ESI/MS in determining accurate loading values and distributions for calicheamicin conjugates are described. Until recently, most MALDI measurements reported in the literature utilized UV lasers. The first reported uses of IR lasers for MALDI/MS, in 1990, demonstrated the feasibility and complementarity of IR-MALDI/MS to UV-MALDI/MS and illustrated a variety of matrix substances for IR-MALDI.28,29 Attractive features of IR-MALDI/MS versus UV-MALDI/MS include a much wider choice of matrix materials that can be used29 and, for macromolecules with molecular weights greater than 50 000, less fragmentation and better resolution.30 Specialized uses of IR-MALDI/MS have been directed to the analysis of blots on membranes,31,32 oligonucleotides,33-35 organometalic compounds,36 and other specialized applications.37 In this study, we were prompted to use IR lasers since UV irradiation decomposes calicheamicin compounds. As will be demonstrated in this report, IR-MALDI mass spectra at 150 000 Da were routinely produced with resolutions up to ∼200, corresponding to better than a 2-fold increase in resolution achieved by UV-MALDI in the same high-mass range. This property of IR-MALDI/MS enabled us to obtain spectra containing information which was used to compute the distribution of the calicheamicin derivatives covalently bound to monoclonal antibodies. EXPERIMENTAL SECTION MoAb and MoAb Conjugate Sample Preparation. The calicheamicin derivatives, referred to as NAc-calicheamicin-DMA (3A, C62H84O24N3S3I, monoisotopic/chemical-averaged MW 1477.4/ 1478.5) and NAc-calicheamicin-AcBut (4A, C73H96O26N5S3I, monoisotopic/chemical-averaged MW 1681.5/1682.5), were prepared as reported elsewhere.38,39 Conjugates of calicheamicin derivative 3A, containing the DMA linker (R2), were designed to be (28) Overberg, A.; Karas, M.; Hillenkamp, F. Rapid Commun. Mass Spectrom. 1991, 5, 128-131. (29) Overberg, A.; Karas, M.; Bahr, U.; Kaufmann, R.; Hillenkamp, F. Rapid Commun. Mass Spectrom. 1990, 4, 293-296. (30) Hillenkamp, F.; Karas, M.; Berkenkamp, S. Proceedings of the 43rd ASMS Conference on Mass Spectrometry and Allied Topics, Atlanta, GA, May 2126, 1995; p 357. (31) Strupat, K.; Karas, M.; Hillenkamp, F.; Eckerskorn, C.; Lottspeich, F. Anal. Chem. 1994, 66, 464-470. (32) Eckerskorn, C.; Strupat, K.; Karas, M.; Hillenkamp, F.; Lottspeich, F. Electrophoresis 1992, 13, 664-665. (33) Nordhoff, E.; Kirpekar, F.; Karas, M.; Cramer, R.; Hahner, S.; Hillenkamp, F.; Kristiansen, K.; Roepstorff, P.; Lezius, A. Nucleic Acids Res. 1994, 22, 2460-2465. (34) Nordhoff, E.; Cramer, R.; Karas, M.; Hillenkamp, F.; Kirpekar, F.; Kristiansen, K.; Roepstorff, P. Nucleic Acids Res. 1993, 21, 3347-3357. (35) Nordhoff, E.; Ingendoh, A.; Cramer, R.; Overberg, A.; Stahl, B.; Karas, M.; Hillenkamp, F.; Crain, P. F. Rapid Commun. Mass Spectrom. 1992, 6, 771776. (36) Costello, C. E.; Nordhoff, E.; Hillenkamp, F. Int. J. Mass Spectrom. Ion Processes 1994, 132, 239-249. (37) Haglund, R. F., Jr.; Tang, K.; Hillenkamp, F.; Chen, C. H. Proc. SPIE-Int. Soc. Opt. Eng. 1993, 1854 (Free-Electron Laser Spectroscopy in Biology, Medicine and Materials Science), 117-128. (38) Ellestad, G. A.; Hamann, P. R.; Zein, N.; Morton, G. O.; Siegel, M. M.; Pastel, M.; Borders, D. B.; McGahren, W. J. Tetrahedron Lett. 1989, 30, 30333036.

Analytical Chemistry, Vol. 69, No. 14, July 15, 1997

2717

Chart 1

hydrolytically stable, while conjugates of 4A, containing the AcBut linker (R2), were designed to hydrolyze under low-pH conditions. The pure hCTM01 and hP67.6 monoclonal antibodies were obtained from Celltech, Ltd. (Slough, Berkshire, UK). The calicheamicin derivatives were covalently linked with primary amines of the MoAbs via amide bonds to form structures 3B,C and 4B,C. Size exclusion chromatography was used to resolve the conjugated calicheamicin derivatives from the unreacted calicheamicin derivatives.39 The salt buffers present in the isolated fractions containing the MoAb conjugates were exchanged into 10 mM ammonium acetate solutions using desalting columns (Pharmacia PD10 or BioRAD 10DG columns). The resultant solutions were diluted to a protein concentration of 1.5 mg/mL and frozen at -20 °C until used. The pure, unconjugated MoAbs were similarly desalted, diluted, and frozen until used. The average loading values for the conjugated MoAbs were determined to range from 1 to 6 mol of calicheamicin derivatives per mole of MoAb using a UV absorption spectrometry method previously described.39 MALDI Mass Spectrometry. The mass spectrometer used for acquiring the IR- and UV-MALDI mass spectral data was a home-built reflector-type time-of-flight mass spectrometer (Mu¨nster). The laser systems used were as follows: for IR, Er-YAG, 2.94 mm (3401 cm-1), pulse width 150-180 ns (electrooptically Q-switched) (Schwartz Electro Optics, Orlando, FL) or 70-100 ns (Spektrum GmbH, Berlin, Germany); for UV, N2 355 nm, pulse width 5 ns (Laser Science, Inc., Newton, MA). The spot size diameters were as follows: for IR, ∼200 µm (focused by a ZnSe lens having a 127 mm focal length) and for UV, 100 µm. The IR and UV laser irradiances were each about 106-107 W/cm2. The acceleration voltage was 12.3 kV. The instrument is equipped (39) Hinman, L. M.; Hamann, P. R.; Wallace, R.; Menendez, A. T.; Durr, F. E.; Upeslacis, J. Cancer Res. 1993, 53, 3336-3342.

2718

Analytical Chemistry, Vol. 69, No. 14, July 15, 1997

with an optical viewing port and microscope with a sample stage which is thermally insulated to make the analysis of cooled samples possible. The matrices used were, for IR, succinic acid (saturated water solution) and frozen pure glycerol, and for UV, 2,5-dihydroxybenzoic acid (saturated solution in 2:1 (v:v) 0.1% TFA/acetonitrile, referred to as DHB-S). The ions were postaccelerated to a potential of 15 kV, detected with a SEM, and recorded with a LeCroy 9400A transient oscilloscope. The spectra acquired consisted of a sum from 10-20 single acquisitions. The data were analyzed with the PC-based ULISSES data evaluation system (Chips At Work, Bonn, Germany) and exported via inhouse software for further data reduction and display with the Macintosh-based KaleidaGraph data analysis/graphics application software (Synergy Software, Reading, PA). The conjugated and pure MoAb samples, originally 1.5 mg/ mL, were prepared for MALDI/MS in the following fashion. Subsamples were dissolved in ultrapure water to a concentration of about 0.10-0.75 mg/mL. Aliquots of the samples (0.5-3.0 µL) were mixed with equal volumes of matrix and deposited on stainless steel targets. The samples with succinic acid and 2,5dihydroxybenzoic acid matrices were air-dried and then inserted into the instrument, while samples with glycerol matrix were cooled with liquid nitrogen and directly inserted into the instrument. The final matrix-to-analyte molar ratio was ∼105 for IRMALDI and ∼106 for UV-MALDI, corresponding to the matrix conditions for optimum signal intensity and resolution. The peak shape function for the molecular ion of a pure MoAb was obtained by optimizing the fit of the sum of three Gaussian functions (sharp, wide, and very wide). The molecular ion envelope of the corresponding conjugated MoAb consisted of a series of peak shape functions similar to that of the pure MoAb, whose peak centers were separated by the mass of the conjugated calicheamicin derivative molecule. Drug distributions were obtained by deconvoluting the molecular ion envelope of the conjugated MoAb using the optimized peak shape function for the conjugated MoAb components. The only unknowns in the deconvolution calculations are the relative intensities of the different peak shape functions making up the distribution of the calicheamicin MoAb conjugates. KaleidaGraph general curve fit algorithms were used to obtain the peak shape functions for the pure and conjugated MoAbs and to deconvolute the molecular ion envelopes of the conjugated MoAbs by computing the best fit of the intensities of the peak shape functions to the experimental data by minimizing the χ2 function. In all the fits, the average relative errors between the fitted and experimental data were less than 1%. The distributions were computed from the relative intensities of the fitted peak shape functions. The average loading values were computed from the sum of the products of the relative intensities of the deconvoluted peaks and the corresponding numbers of conjugated drugs, normalized by the sum of the relative intensities. Electrospray Ionization Mass Spectrometry. Electrospray ionization mass spectra were obtained with a Micromass Quattro triple-quadrupole mass spectrometer equipped with a Micromass electrospray source, rf hexapole lens, and Megaflow gas nebulizer probe. The capillary sprayer voltage was set to ∼3.0 kV, the highvoltage lens set to 250 V, the nozzle-skimmer (sampling cone) voltage maintained at ∼20 V (or ∼95 V) for wide (or narrow) charge distributions, and lens 3 set to -90 V. The source temperature was maintained at ∼80 °C. The nebulizer and bath

gases were both nitrogen, delivered at flow rates of 15 and 500 L/h. The mass spectrometer was set to unit resolution and calibrated with aqueous cesium iodide solutions. The desalted aqueous MoAb samples were prepared to a concentration of ∼10 pmol/µL in a solvent consisting of ∼35% acetonitrile and 3% acetic acid. About 10 µL of each MoAb sample was “sandwiched” on either side with 5 µL of 3% acetic acid dissolved in 1:1 acetonitrile/ water. The solvent sandwich was utilized to minimize peak broadening due to binding of the protein to the tubing walls. The MoAb samples were infused into the source of the mass spectrometer at a flow rate of ∼4 µL/min utilizing a carrier solvent of 1:1 acetonitrile/water with either an ISCO µLC-500 syringe pump or an ABI Model 140B dual-syringe pump. Data were acquired at 16 data points/Da over the wide mass range of ∼1900-3200 Da or over the narrow mass range of ∼2475-2650 Da, with scan times of from 30 to 45 s. Ten to twenty spectra were averaged, smoothed, baseline subtracted, centroided, and transformed from a mass/charge (m/z) axis to a mass axis. The charge and mass assignments for related spectral peaks were assigned by minimizing the average mass error in the calculated molecular weight by varying systematically the possible charge assignments for the peaks in the envelope. The Micromasssupplied electrospray maximum entropy algorithm40 was also used to transform the spectra to a mass axis and to resolution enhance the transformed spectra. The maximum entropy algorithm was set to optimize the spectra at peak widths of 2 Da at m/z ∼2500, which corresponds to predicted natural peak widths of ∼40 Da for MoAbs with molecular masses of 150 000 Da. The resulting resolution-enhanced spectral peaks were then integrated to display the correct ion abundances for distribution analysis.41 The resulting charge and mass assignments from the transform and maximum entropy algorithms were found to be consistent. RESULTS Degradation of Calicheamicin Derivatives and NAc-Calicheamicin-MoAb Conjugates by UV Irradiation. Under IRMALDI/MS conditions, the base peak in the spectrum of calicheamicin derivative 3A is the natriated molecular ion, [M + Na]+, at m/z 1500 (Figure 1A). The natriated molecular ion mass is confirmed by the less intense ion at m/z 1516, corresponding to [M + K]+. The weak ions at m/z 1322 and 1309 are the only higher mass components that may have arisen as a result of decomposition upon IR irradiation and account for about 20% of the total ion current in the mass range of the molecular ion. The ion with m/z 1322 is a natriated molecular ion (vide infra) with a MW of 1299. This molecule is probably a degradation component, and it is 34 Da (consistent with H2S) lower in mass than NAccalicheamicin  (2), suggesting a related structure, perhaps one with an alternative mode of cyclization, although no satisfactory structure has been assigned. The IR-MALDI mass spectrum of calicheamicin derivative 3A after irradiation with a low-intensity UV lamp outside of the mass spectrometer (Figure 1B) exhibited no molecular ions but rather degradation ions at m/z 1322 and 1338 corresponding to sodium and potassium adduct ions, respectively, of the degradation component with MW 1299. Under the high-energy irradiation of the UV-MALDI/MS laser, cali(40) Ferrige, A. G.; Seddon, M. J.; Green, B. N.; Jarvis, S. A.; Skilling, J. Rapid Commun. Mass Spectrom. 1992, 6, 707-711. (41) Cottrell, J. C.; Green, B. N. Proceedings of the 41st ASMS Conference on Mass Spectrometry and Allied Topics, San Francisco, CA, May 31-June 4, 1993; pp 146a-b.

Figure 1. Mass spectra of NAc-calicheamicin-DMA (3A, monoisotopic MW 1477.4) produced by (A) IR-MALDI/MS with a pulsed IR laser (matrix, succinic acid), (B) IR-MALDI/MS after the sample was irradiated with UV radiation outside the mass spectrometer (matrix, succinic acid), and (C) UV-MALDI/MS with a pulsed UV laser (matrix, dihydroxybenzoic acid). (D) IR-MALDI mass spectrum of NAccalicheamicin-AcBut (4A, monoisotopic MW 1681.5) (matrix, glycerol). The rise in the baselines of the spectra is indicative of the extent of fragmentation and degradation of the samples.

cheamicin derivative 3A produced a rather weak [M + Na]+ and an abundant natriated molecular ion of the calicheamicin degradation product with MW 1299 (Figure 1C). In a similar fashion, NAc-calicheamicin-AcBut (4A) in the IR-MALDI/MS mode produced abundant [M + Na]+ and [M + K]+ ions (Figure 1D), while in the UV-MALDI/MS mode the natriated molecular ion of 4A was of very low abundance and the natriated molecular ion at m/z 1322 of the proposd degradation component was very abundant (data not shown). These experiments demonstrate that calicheamicin derivatives are relatively stable when exposed to the IR laser irradiation energies used in IR-MALDI/MS but, when exposed to UV radiation, readily undergo photochemical degradation, consequently producing molecular ions of only very low abundances in the UV-MALDI/MS mode. The effect of increasing UV exposure time on the integrity of NAc-calicheamicin-DMA-hCTM01 MoAb conjugate (3B) was studied by exposing the sample to UV radiation from a lowintensity UV lamp or to sunlight and analyzing the resultant Analytical Chemistry, Vol. 69, No. 14, July 15, 1997

2719

Figure 2. IR-MALDI experimental and mathematically fitted peak profiles of singly and doubly charged molecular ion envelopes, respectively: (A-1 and A-2) pure hCTM01 MoAb and (B-1 and B-2) normal loaded sample of NAc-calicheamicin-DMA-hCTM01 MoAb conjugate (3B, UV absorption spectrometry loading of 2.00). The insets to A-1 and A-2 are overlays of the respective IR- and UV-MALDI peak profiles demonstrating the enhanced resolution achieved with the IR versus the UV laser. For B-1 and B-2, note the partial resolution of the molecular ion envelopes and the deconvoluted peak profiles for each of the calicheamicin conjugates contributing to the drug distribution. Also, included with each figure is the difference in relative intensity (error) between the experimental and fitted data. The matrix used for these IR-MALDI spectra was frozen glycerol.

products by IR-MALDI/MS (vide infra). As illustrated in Supplementary Figure 1, the mass centroid of the conjugate 3B decreases with increasing exposure time to UV irradiation, while the mass centroids for pure hCTM01 and the peak maxima for the component of 3B corresponding to pure hCTM01 maintained the same value during all the irradiation experiments, within experimental precision of about (125 Da. Also, the peak shape on the low-mass side of this component of 3B was absolutely unaffected during the whole exposure time of 200 min, indicating that no significant degradation occurred to this pure hCTM01 component. Note that, at long exposure times, the mass of the conjugate is nearly identical to that of pure hCTM01 MoAb. Presumably, the conjugated calicheamicin upon UV irradiation undergoes a similar photochemical degradation to that described above for the formation of the ion at m/z 1322, which apparently decouples the calicheamicin derivative from the MoAb by effectively cleaving the linkage between the calicheamicin and the MoAb. These experiments demonstrate the instability to UV irradiation of conjugated calicheamicin derivatives linked by 2720

Analytical Chemistry, Vol. 69, No. 14, July 15, 1997

disulfide bonds to the MoAb and implicitly demonstrates the relative stability of these molecules to IR irradiation. IR-MALDI Mass Spectral Characteristics of Pure MoAbs and NAc-Calicheamicin-MoAb Conjugates. In terms of information content, the IR-MALDI mass spectra of pure monoclonal antibodies are comparable to those produced by UVMALDI. The IR-MALDI mass spectra for pure hCTM01 and hP67.6 MoAbs using frozen glycerol and succinic acid as the matrices exhibited abundant M+, M2+, and M3+ ions and generally less abundant dimer ions 2M+ and 2M3+ and the trimer ion 3M2+. However, polymer formation and the distribution of the charged species generally is dependent on the matrix-to-analyte molar ratio used during sample preparation. In the IR-MALDI mass spectrum, when using glycerol matrix, no ions related to the light or heavy chains of the MoAb were observed, while with succinic acid matrix, low abundances were observed, generally with less than 10% relative intensity. Parts A-1 and A-2 of Figure 2 respectively illustrate the singly and doubly charged parent ion peaks for pure hCTM01 MoAb using (frozen) glycerol as the matrix. Similar

Figure 3. IR-MALDI experimental and mathematically fitted peak profiles of singly charged molecular ion envelopes: (A) pure hP67.6 MoAb and (B) high loaded NAc-calicheamicin-AcBut-hP67.6 MoAb conjugate (4B, UV absorption spectrometry loading of 5.52). For (B), note the partial resolution of the molecular ion envelopes and the deconvoluted peak profiles for each of the calicheamicin conjugates contributing to the drug distribution. Also, included with the figure is the difference in relative intensity (error) between the experimental and fitted data. The matrix used for these IR-MALDI spectra was frozen glycerol.

results were obtained using succinic acid as the matrix.10 The peak maxima and peak shapes were identical for both matrices. Both samples exhibited some tailing on the high-mass sides of the peaks, which was attributed to matrix adduct formation. Similar results were obtained with hP67.6 MoAb. Figure 3A illustrates the singly charged molecular ion peak for pure hP67.6 using (frozen) glycerol as the matrix. The IR-MALDI spectra are highly reproducible. Mass centroids for the singly and doubly charged MoAb molecular ions are reproducible, with average mass errors less than (0.1%. An unusual feature of the MALDI mass spectra of the MoAbs is that resolutions up to ∼200 are achieved routinely in the IRMALDI mode, which is about 2 times greater than can be achieved in equivalent measurements in the UV-MALDI mode. The singly and doubly charged molecular ion envelopes for pure hCTM01 MoAb, acquired in both the UV- and IR-MALDI/MS modes, are illustrated respectively in Figure 2, insets to parts A-1 and A-2.

Note that the bandwidth at half-maximum (fwhm) for the IRMALDI peak is about half the width of the superimposed UVMALDI peak. A clear distinction between the UV- and IR-MALDI spectra is the broad tailing on the low-mass side of the peak produced in the UV mode versus the sharp rise and absence of such tailing in the IR-MALDI mode. Supplementary Table 1 lists the measured peak widths and resolutions for the singly and multiply charged hCTM01 MoAb parent ions in the UV- and IRMALDI/MS modes. These results demonstrate that resolution enhancement in the IR-MALDI mode can be achieved with glycerol and succinic acid matrices for singly and multiply charged materials with molecular masses of 150 000 Da. Recently, Hillenkamp et al.30 demonstrated resolution and sensitivity enhancement in IR-MALDI/MS by use of glycerol at subambient temperatures. The mechanism for resolution enhancement in the IRMALDI mode is actively under investigation. When IR-MALDI/MS was applied to MoAbs conjugated to calicheamicin derivatives, all the peak profiles in the observed spectra are broadened relative to the pure MoAb peak widths and exhibit a series of partially resolved peaks. The degree to which the peaks are broadened is a function of the amount of loading of the calicheamicin derivatives to the MoAb. The mass shift difference between adjacent peak maxima corresponds to the mass of one calicheamicin derivative molecule covalently attached to the MoAb. This mass shift arises as a result of the additional resolution available in the IR-MALDI experiment which is sufficient to partially resolve the distribution of calicheamicin components which are covalently attached to the MoAb molecules. Parts B-1 and B-2 of Figure 2, respectively, illustrate the singly and doubly charged parent ion peaks for normal levels of NAccalicheamicin-DMA conjugated to hCTM01 MoAb (3B) using (frozen) glycerol as the matrix. Similar results were obtained using succinic acid as the matrix.10 The mass difference between adjacent maxima of the mass shifted peaks is 1460 Da, within experimental error, corresponding to the mass change for the covalent addition of NAc-calicheamicin-DMA molecules (3A) to the MoAb. Figure 3B illustrates the singly charged molecular ion peak for large amounts of calicheamicin-AcBut covalently attached to hP67.6 MoAb (4B) using (frozen) glycerol as the matrix. The maximums in the partially resolved peaks are 1664 Da apart, within experimental error, corresponding to the mass change for the addition of NAc-calicheamicin-AcBut molecules (4A) to the antibody. Determination of Average Loading Values and Distributions for Calicheamicin Derivatives Conjugated to MoAbs from IR-MALDI Data. As described in the Experimental Section, the distributions of the calicheamicin derivatives conjugated to the MoAbs were calculated by deconvoluting the partially resolved IR-MALDI molecular ion envelopes for the MoAb conjugates using the optimized peak shape function of the conjugated MoAb components. Supplementary Table 2 summarizes, for NAc-calicheamicin-DMA-hCTM01 MoAb conjugates (3B), the calculated distributions for samples with UV absorption spectrometry average loadings of 2.00 (illustrated in parts B-1 and B-2 of Figure 2 using a glycerol matrix) and 2.70 (not illustrated). The corresponding IR-MALDI/MS average loading values for the singly charged molecular ions were 2.05 and 2.89, respectively. It is clear that, for these conjugates made with the hydrolytically stable DMA linker between calicheamicin and the MoAb, the average loading values computed from the IR-MALDI/MS data Analytical Chemistry, Vol. 69, No. 14, July 15, 1997

2721

Table 1. Predicted and Experimental Loading Values for NAc-Calicheamicin-DMA-hCTM01 (3B), Titrated into Pure hCTM01 molar ratio curvea

pure hCTM01

A B C D E

1.00 1.00 1.00 1.00 1.00

loading value 3B

predictedb

observed

error (%)

5.00 4.00 3.00 2.00 1.00

2.34 2.25 2.11 1.87 1.40

2.58 2.49 2.33 1.88 1.33

-10.1 -10.9 -10.4 -0.4 5.7

a The deconvoluted titration curves are illustrated in Supplementary Figure 2. b The predicted values were computed on the basis of the IR-MALDI/MS loading value of 2.80 for a freshly prepared sample of NAc-calicheamicin-DMA-hCTM01 (3B).

were very similar to those computed from UV absorption spectrometry data. This was not true for the UV-MALDI/MS data previously reported.3 Two trends were observed in the data described in the paragraph above. One trend is related to the IR-MALDI/MS average loading values obtained with the glycerol and succinic acid matrices. Despite the fact that the resolution and spectral quality for spectra obtained with glycerol and succinic acid matrices are almost the same, the loading values obtained with the glycerol matrix tend to be closer to the UV absorption values, while the loading values obtained with succinic acid were, on average, about 20% lower than those obtained with the glycerol matrix. Since the peak shapes were similar for glycerol and succinic acid, perhaps some of the conjugated material is lost with succinic acid, as indicated by greater abundances of the members in the distributions containing 0, 1, 2, and 3 drug molecules (data not shown) versus the same parameters for glycerol. Glycerol may be a more universal matrix than succinic acid in the sense that degradation and hydrolysis should be suppressed since lowpH effects are absent. The second trend is that the IR-MALDI/ MS average loading values, calculated from distributions, appear to be consistent with the average loading values obtained by UV absorption spectrometry. The advantage of the distribution average loading value is that it is independent of drift in the instrument calibration since it is calculated from the deconvoluted distribution of a single molecular ion envelope measurement. The major assumptions that should be kept in mind with the distribution loading value calculations are that the peak shape, relative amount of matrix adduct formation, and instrumental response for each member in the distribution are presumed to be identical. Titration experiments were performed in which NAc-calicheamicin-DMA-hCTM01 (3B) was titrated into a fixed amount of pure hCTM01 to demonstrate the reliability of the IR-MALDI/ MS experiments and, thereby, the reliability of the experimentally determined IR-MALDI/MS loading values. A series of IR-MALDI/ MS experiments were performed in which the molar ratios of pure hCTM01 to 3B were 1:1, 1:2, 1:3, 1:4, and 1:5. The predicted and observed IR-MALDI/MS loading values are listed in Table 1. These controlled experiments demonstrate that small changes in loading values can be accurately determined, with the poorest fits off by ∼10%. Supplementary Figure 2 illustrates the IR-MALDI mass spectra for these samples deconvoluted into the same fixed amount of pure hCTM01 and the hCTM01 conjugate (3B) contributions. Note that the peak (labeled F) corresponding to the fixed amount of pure hCTM01 is normalized to the same 2722

Analytical Chemistry, Vol. 69, No. 14, July 15, 1997

intensity for each of the samples, while the relative abundance of 3B linearly increases with increasing ratios of 3B (peaks labeled A-E). Also note that the hCTM01 conjugate (3B) distributions contain shoulders on the low-mass side of the peak envelopes, corresponding to no conjugated drug, i.e., pure hCTM01. Significantly different results were obtained for NAc-calicheamicin-AcBut-hP67.6 (4B) when the conjugate was analyzed by IR-MALDI/MS. For normal loaded samples, UV absorption spectrometry predicted a loading of ∼2.8, but the samples were found to have a loading by IR-MALDI/MS (using frozen glycerol as the matrix) of only ∼0.3. Similarly, a sample of NAccalicheamicin-AcBut-CTM01 MoAb conjugate (4C), prepared with a loading of ∼3.7 by UV absorption, produced an IR-MALDI/MS loading value of ∼0.3. On the other hand, the IR-MALDI mass spectrum (using frozen glycerol as the matrix) for a high loaded (UV absorption loading of ∼5.5) NAc-calicheamicin-AcBut-hP67.6 MoAb conjugate (4B), illustrated in Figure 3B, had a bimodial distribution consisting of a very sharp and abundant distribution of lower loading and a very broad distribution of higher loading. Listed in Supplementary Table 3 are the IR-MALDI/MS drug distribution and average loading value (5.54). Although there is apparent agreement between the UV absorption loading value (5.52) and the IR-MALDI/MS loading, the large amounts of MoAb carrying 0 or 1 NAc-calicheamicin-AcBut molecules leads one to strongly suspect significant loss of drug from the hP67.6 MoAb. A Gaussian distribution was expected for the high loaded AcBut MoAb conjugate 4B rather than the bimodial distribution with contributions from low and high loading regions. The anomolous results with the normal and high loaded conjugates with the AcBut labile linkers 4B,C are presumably related to the chemical instability of the linker under the experimental conditions, such that highly loaded materials partially degrade while normal loaded materials degrade extensively. It may be that normal and high loaded samples with the AcBut linker need to be protected from degradation during analysis. Techniques to overcome this problem are being investigated. Note, however, that ESI/MS results (described below) were qualitatively consistent with the IR-MALDI/MS findings, even though the ESI/MS experimental conditions were quite different from the MALDI/MS conditions. ESI Mass Spectral Characteristics of Calicheamicin Derivatives 3A and 4A. The calicheamicin derivatives 3A and 4A (dissolved in acetonitrile with 3% acetic acid) were analyzed by ESI/MS and produced the spectra illustrated in Figure 4. Both spectra exhibit abundant molecular ions of the type [M + H]+ and less abundant doubly charged molecular ions of the type [M + 2H]2+, [M + H + Na]2+, and [M + H + K]2+. The only significant fragment ions produced under the ESI conditions were related to the loss of the amino sugar from the calicheamicin derivatives, viz., the abundant ions at m/z 200, corresponding to the protonated amino sugar, and the less abundant ions at m/z 1279 and 1483 for compounds 3A and 4A, respectively, corresponding to the loss of the amino sugar from the protonated molecular ions. Other minor fragment ions were identified and are indicated in Scheme 1 for the calicheamicin derivative structures 3A and 4A. Essentially, the abundance of these observed fragment ions for the calicheamicin derivatives is expected to be significantly reduced when these small molecules are conjugated to much more massive MoAbs. The chemical instability of the NAc-calicheamicin-AcBut (4A) is graphically

Figure 4. ESI mass spectra of (A) NAc-calicheamicin-DMA (3A, monoisotopic MW 1477.4), and (B) NAc-calicheamicin-AcBut (4A, monoisotopic MW 1681.5). The solvent background (3% acetic acid in acetonitrile) was subtracted from each spectrum. The proposed fragment ion structures common for both compounds are illustrated in Scheme 1.

Scheme 1. ESI/MS of NAc-Calicheamicin-DMA (3A) and Calicheamicin-AcBut (4A)

illustrated in Figure 4B by the ions at m/z 1478 and 1279, corresponding to the protonated molecular ion for the calicheamicin hydrolysis product of the hydrazone bond and the loss of the the amino sugar from the calicheamicin hydrazone product,

respectively. For comparison purposes, the ESI/MS experimental conditions used for the calicheamicin derivatives were identical to the more energetic conditions (nozzle-skimmer voltage 20 V) used for the calicheamicin MoAb conjugates. Analytical Chemistry, Vol. 69, No. 14, July 15, 1997

2723

Figure 5. Wide-scan ESI mass spectra of (A) pure hCTM01 MoAb and (B) lower loaded calicheamicin-DMA-hCTM01 MoAb conjugate (3B, UV absorption spectrometry loading of 2.00).

ESI Mass Spectral Characteristics of Pure MoAbs and NAc-Calicheamicin-MoAb Conjugates and the Determination of Average Loading Values and Distributions for the Calicheamicin Conjugates from the ESI Mass Spectral Data. Wide- and narrow-scan ESI/MS spectra for pure hCTM01 MoAb are illustrated in Figures 5A and 6A-1, respectively. Unique features of the spectra are the broad charge distribution and partial resolution of the peaks. No ions related to the light or heavy chains of the MoAb were observed. The transformed and integrated maximum entropy (resolution enhanced) spectra are illustrated in parts A-2 and A-3, respectively, of Figure 6 for the narrow-scan data. The calculated molecular weight for hCTM01 is 149 627 + 162n, where n ) 0, 1, 2, 3, and 4. The partial resolution of the molecular ion peak is due to the heterogeneity of the polysaccharide present in the MoAb which contains resolvable components, with mass increments of 162 Da, consistent with the addition of mannose or galactose units to the lowest mass immunoglobulin glycan. Recently, complicated glycosylation patterns have been reported for the humanized CAMPATH 1H MoAb42 and MoAb fragments.43,44 The transformed ESI mass spectrum of the reaction products of hCTM01 MoAb treated with dithiothreitol consisted of a pure light chain with MW 24 130 and a mixture of heavy-chain products due to the glycan heterogeneity with MWs 50 705 + 162n, where n ) 0, 1, and 2 (see Supplementary Figure 3A). Lower and higher loaded NAc-calicheamicin-DMA-hCTM01 MoAb conjugates 3B, with UV absorption loading values of 2.00 and 2.70, respectively, were analyzed by ESI/MS. (The corresponding results obtained by IR-MALDI/MS were described above.) Wide- and narrow-scan ESI/MS spectra for the lower (42) Ashton, D. S.; Beddell, C. R.; Cooper, D. J.; Craig, S. J.; Lines, A. C.; Oliver, R. W. A.; Smith, M. A. Anal. Chem. 1995, 67, 835-842. (43) Bourell, J. H.; Clauser, K. P.; Kelley, R.; Carter, P.; Stults, J. T. Anal. Chem. 1994, 66, 2088-2095. (44) Jiskoot, W.; van de Werken, G.; Coco Martin, J. M.; Green, B. N.; Beuvery, E. C.; Crommelin, D. J. A. Pharm. Res. 1991, 9, 945-951.

2724

Analytical Chemistry, Vol. 69, No. 14, July 15, 1997

loaded sample are illustrated in Figures 5B and 6B-1 (higher loaded sample not shown), respectively. Besides the broad charge distribution and partial resolution of the peaks which appeared in the ESI spectra of pure hCTM01 MoAb (Figures 5A and 6A1), two new features appear. One glaring feature is a new series of partially resolved peaks with a broad charge distribution between the peaks, corresponding to the pure hCTM01 MoAb. A second, more subtle feature is the higher abundances on the high-mass sides of the series of peaks corresponding to the pure hCTM01 MoAb. These abundances are significantly greater for the higher than the lower loaded MoAb samples. The transformed and integrated maximum entropy data for the lower loaded 3B conjugate in the narrow-scan mode are illustrated in parts B-2 and B-3 of Figure 6 (higher loaded sample not shown), respectively. Note the appearance of a new series of peaks, each ∼1460 Da apart, relative to the peaks in the corresponding spectra for pure hCTM01 MoAb (Figure 6, parts A-2 and A-3). This mass difference corresponds to the mass change of the MoAb after the covalent addition of NAc-calicheamicin-DMA (3A) via an amidation reaction. From the series of observed ions, one concludes that a distribution of from one to four NAc-calicheamicin-DMA molecules conjugated with hCTM01 MoAb is present in the samples. The qualitative distribution of the ions for the lower and higher loaded samples only differs by the amount of nonconjugated hCTM01 MoAb, while the distribution of loaded MoAb is essentially the same for both samples. These observations are consistent with the way the samples were prepared, namely, pure hCTM01 was titrated into the higher loaded MoAb to create the lower loaded sample. An expected feature for each of the conjugates is a glycan distribution similar to that of the pure hCTM01. This was not observed because of the overlap of the multiply charged peaks associated with the distributions of the glycans and the calicheamicin conjugates. Listed in Supplementary Table 2 are the calculated distributions for the lower and higher loaded samples from the integrated maximum entropy

Figure 6. Narrow-scan ESI mass spectra: (A-1, -2, -3) raw, transformed, and integrated maximum entropy data for pure hCTM01 MoAb, respectively, and (B-1, -2, -3) raw, transformed, and integrated maximum entropy data for lower loaded NAc-calicheamicin-DMA-hCTM01 MoAb conjugate (UV absorption spectrometry loading of 2.00), respectively.

values from the ESI/MS data illustrated in Figure 6B-3 for the lower loaded sample (higher loaded sample not shown). Also included in Supplementary Table 2 are the average loading values for the respective samples computed from the drug distribution values. These ESI/MS distribution average loading values are lower than the UV absorption average loading values and less reliable than the IR-MALDI/MS results due to the difficulty in resolving the overlapping, multiply charged peaks arising from the heterogeneity of the covalently bound NAc-calicheamicin-DMA drug and the polysaccharide in the monoclonal antibody. The ESI mass spectrum of pure hP67.6 MoAb exhibited partially resolved peaks, each with a mass shift computed to be 162 Da. The spectrum was very similar in appearance to the spectrum obtained for hCTM01 MoAb (Figures 5A and 6A-1). No ions related to the light or heavy chains of the MoAb were observed. The measured molecular weight for hP67.6 MoAb is 147 783 + 162n, where n ) 0, 1, 2, and 3 and corresponds to the number of resolvable polysaccharide units, each differing by 162 Da. The molecular mass of hP67.6 is about 2 kDa less than the molecular mass of hCTM01 MoAb. The transformed ESI mass spectrum of the reaction products of hP67.6 MoAb treated with dithiothreitol consisted of a pure light chain with MW 23 821 and a mixture of heavy-chain products due to the glycan heterogeneity with MWs 50 106 + 162n, where n ) 0, 1, and 2 (see Supplementary Figure 3B). For comparison purposes, the transformed ESI mass spectrum of dithiothreitol-treated deglycosylated hP67.6 MoAb is illustrated in Supplementary Figure 3C, demonstrating that the mass of the carbohydrate content in hP67.6 MoAb is 2878 + 162n, where n ) 0, ..., 4.

Figure 7. Transformed ESI mass spectra for normal loaded NAccalicheamicin-AcBut-hP67.6 MoAb conjugate (4B, UV absorption spectrometry loading of ∼2.8).

NAc-calicheamicin-AcBut-hP67.6 MoAb conjugate (4B), with a UV absorption loading value of ∼2.8, was analyzed by ESI/MS and found to have a loading of only ∼0.3 (see Figure 7). Similarly, a sample of NAc-calicheamicin-AcBut-hCTM01 MoAb conjugate (4C), prepared with a loading of ∼3.7 by UV absorption, was found to have an ESI/MS loading value of ∼0.3 (not shown). These low ESI/MS loading values were consistent with the low IRMALDI/MS loading values (described above) and are probably due to the chemical instability of the linker under the experimental conditions. On the other hand, the high loaded NAc-calicheamiAnalytical Chemistry, Vol. 69, No. 14, July 15, 1997

2725

cin-AcBut-hP67.6 MoAb conjugate (4B), analyzed above by IRMALDI/MS (illustrated in Figure 3B and summarized in Supplementary Table 3), having a loading value of 5.5 by UV absorption, produced by ESI/MS a broadened, partially resolved spectrum. This ESI mass spectrum is consistent with a highly loaded sample where heterogeneity of charge and sample components (polysaccharides and NAc-calicheamicin-AcBut conjugates) overlap to produce a broadened, partially resolved continuum. CONCLUSIONS IR-MALDI/MS with glycerol matrix has been found herein to be a reliable method for determining the average loading values and distributions of calicheamicin derivatives conjugated to MoAbs with hydrolytically stable linkers. However, measurements made on NAc-calicheamicin-MoAb conjugates by IR-MALDI/MS with succinic acid matrix, by ESI/MS and by UV-MALDI/MS, did not produce useful distribution values or loadings that correlated with the measured UV absorption spectrometry loading values. Hydrolysis of the linkage between drug and MoAb at low pH may be responsible for the low loading values when succinic acid is used as the matrix in IR-MALDI/MS and acetic acid is used as the ionizing solvent in ESI/MS. The overlap of the multiply charged peaks associated with the different oligomeric species present in the mixture and the possibility of chemical and electrochemical degradation processes occurring in the source probably are responsible for the problems with the ESI/MS measurements. UV-MALDI/MS failed, presumably because the conjugated calicheamicin derivatives are photochemically unstable when exposed to UV radiation, as is the drug alone. In spite of the success in applying IR-MALDI/MS with frozen glycerol matrix, a number of areas deserve additional attention in future work. Since the UV absorption loading values for all the reported samples have been confirmed by biochemical and immunological methods, it is clear that, even with the improvements reported in this paper, there is a small amount of degradation occurring during IR-MALDI/MS. This is most apparent with normal loaded hydrolytically unstable conjugates. Matrices other than frozen glycerol that could serve to protect the drug during analyses need to be found. Additionally, there

2726

Analytical Chemistry, Vol. 69, No. 14, July 15, 1997

may be pretreatments of the conjugate (e.g., refolding of the protein) that will cause the protein itself to protect the drug from degradative processes during mass spectral analyses. Further advances in any of these areas would make MALDI/MS into a standard tool for determining loadings and distributions of all types of stable and unstable small molecules conjugated to proteins. Finally, the achievement of higher mass resolution and higher sensitivity in IR-MALDI/MS and ESI/MS should yield even more reliable average loading and distribution values. ACKNOWLEDGMENT The authors greatly appreciate the initial exploratory MALDI/ MS work of Arnd Ingendoh and Michael Karas and the advice and counsel of Janis Upeslacis, Dan Shochat, and Martin P. Kunstmann in supporting this work. SUPPORTING INFORMATION AVAILABLE Three tables, showing mass resolution comparison for UV- and IR-MALDI/MS with hCTM01 monoclonal antibody and relative intensities (distribution) of deconvoluted molecular ion peaks and average loading values computed from the IR-MALDI and ESI mass spectra of 3B and 4B; three figures, showing mass centroids (50%) as a function of increasing exposure times to UV irradiation as measured by IR-MALDI/MS of NAc-calicheamicin-DMA conjugated to hCTM01 MoAb, 3B, pure hCTM01 MoAb, and the peak maxima for the component of 3B corresponding to pure hCTM01, deconvoluted IR-MALDI mass spectra for NAc-calicheamicin-DMA-hCTM01 3B (using frozen glycerol as the matrix) titrated into a fixed amount of pure hCTM01, and transformed ESI mass spectra of the light-chain and heavy-chain regions of dithiothreitol-treated hCTM01 MoAb, hP67.6 MoAb, and deglycosylated hP67.6 MoAb (7 pages). Ordering information is given on any current masthead page. Received for review January 13, 1997. Accepted April 23, 1997.X AC970035Q X

Abstract published in Advance ACS Abstracts, June 15, 1997.