Bioconjugate Chem. 1997, 8, 391−399
391
Structural Determination of the Conjugate of Human Serum Albumin with a Mitomycin C Derivative, KW-2149, by Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry Tohru Yasuzawa and Kenneth B. Tomer* Laboratory of Structural Biology, National Institute of Environmental Health Sciences, P.O. Box 12233, Research Triangle Park, North Carolina 27709. Received December 9, 1996X
A new mitomycin C derivative, KW-2149, is known to form a covalent conjugate with human serum albumin (HSA). This conjugate exhibits 1/20 of the anticellular activity of unconjugated KW-2149. Structural studies of this conjugate were carried out using a combination of enzymatic digestion, highperformance liquid chromatography (HPLC), and matrix-assisted laser desorption/ionization (MALDI) mass spectrometry. The tryptic peptide T5 (residues 21-41) was the only peptide found to be modified by KW-2149 moieties, the [(γ-L-glutamylamino)ethyl]thio group or the (2-aminoethyl)thio group, through a disulfide bond. Although the latter peptide lost its mitomycin C moiety in the course of tryptic digestion, these data strongly suggest that KW-2149 was bound to Cys-34, the only free cysteine on HSA.
INTRODUCTION
Drug-protein binding is one of the essential factors in analyzing pharmacokinetics because it can affect the distribution, metabolism, and excretion of a drug within the body. It is also known that one drug can compete with a second drug for the drug-binding site on serum proteins in man which is the case with clofibrate and warfarin (1). This drug substitution phenomenon can alter the drug concentration levels in blood, causing side effects in some cases. For this reason, identifying the drug-binding site on serum proteins of each drug is necessary for developing new drugs and using them safely. A large number of small molecules, including many drugs, bind reversibly to serum proteins, which act as circulatory transporters (2). The most abundant protein in the human circulatory system is serum albumin (HSA). The protein (HSA, Mr ) 65K) consists of a single, nonglycosylated, polypeptide containing 585 amino acids (Figure 1), one free thiol (Cys-34), and a total of 17 disulfide bridges (3, 4) The drug-binding sites of HSA have been well studied, and two major binding sites have been identified (5-9). It is known that some compounds, such as penicillin, acyl glucuronide, and glucose, bind covalently to HSA via transacylation or via the formation of a Schiff base at lysine or tyrosine residues (10-15). The cysteine residue is also a covalent binding site, to which biological thiols bind through a disulfide linkage (2). Recently, a new mitomycin C derivative, KW-2149 (Figure 2), has been reported (16-20). This compound exhibits a broad spectrum antitumor activity in experimental tumor models, including mitomycin C-resistant tumors. It is also less hematotoxic than mitomycin C. The metabolic half-life of KW-2149 is very rapid with a t1/2 of 9.7 min. One of the major metabolites of KW-2149 has been identified as the albumin conjugate (18). This conjugate still has 1/20 of the anticellular activity of * Corresponding author. Phone: (919) 541-1966. Fax: (919) 541-7880. E-mail:
[email protected]. X Abstract published in Advance ACS Abstracts, May 1, 1997.
S1043-1802(97)00038-4 CCC: $14.00
Figure 1. Amino acid sequence of HSA deduced from cDNA sequences by Putnam et al. (3). Underlined tryptic fragment labels indicate that the tryptic fragment was observed by MALDI-MS.
A
B
Figure 2. Chemical structure of KW-2149. This compound is composed of two sulfides, γ-L-glutamylcysteamine and 7-N-(2mercaptoethyl)mitomycin C. These two parts are described as left-half (A) and right-half (B) sulfides of KW-2149, respectively, in this paper.
unconjugated KW-2149. Because KW-2149 contains a disulfide bond, it is proposed that it binds to HSA at a © 1997 American Chemical Society
392 Bioconjugate Chem., Vol. 8, No. 3, 1997
cysteine residue through formation of a disulfide linkage. There is a great interest in the structure of this conjugate in view of not only the drug-protein binding but also the activation mechanism of this mitomycin C-HSA disulfide. In the study of the activation mechanism of KW2149, thiol molecules such as glutathione and cysteine were found to significantly enhance the cytotoxicity of this drug possibly through reduction of the disulfide moiety of KW-2149 to a sulfur ion which in turn activates the quinone ring to the corresponding semiquinone by an intramolecular reaction (21). The albumin conjugate of KW-2149 may be activated by the same mechanism. This proposed activation mechanism differs from that of mitomycin C which proceeds through reductive activation by DT-diaphorase and cytochrome P450 reductase (22, 23). The difference between the two activation mechanisms is thought to be the reason why KW-2149 is effective against mitomycin C-resistant tumors. The formation of a conjugate with HSA which still has potent anticellular activity is one of the characteristic features of KW-2149. Understanding this phenomenon, and especially identifying the structure of the conjugate, are of great interest in the analysis of the pharmacokinetics of KW-2149, as well as in the study of the activation mechanisms of this drug. We have undertaken a study to determine the structure of the conjugate, including the position of attachment between KW-2149 and HSA, using a combination of enzymatic digestion, high-performance liquid chromatography (HPLC) separation, and matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS). EXPERIMENTAL PROCEDURES
Chemicals and Materials. Human serum albumin (HSA), fraction V, fatty acid free, was purchased from Bayer Corp. (Kankakee, IL). Sequencing grade trypsin and carboxypeptidase P were purchased from Boehringer Mannheim GmbH (Indianapolis, IN). 7-N-[2-[[2-(γ-LGlutamylamino)ethyl]dithio]ethyl]mitomycin C (KW2149) was synthesized in the Pharmaceutical Research Laboratories, Kyowa Hakko Kogyo Co., Ltd. (Shizuoka, Japan). Acetonitrile, obtained from Fisher Scientific (Pittsburgh, PA), water, obtained from J. T. Baker (Phillipsburg, NJ), and trifluoroacetic acid (TFA), obtained from Pierce (Rockford, IL), were HPLC grade. Dithiothreitol (DTT) and R-cyano-4-hydroxycinnamic acid (CHCA) were purchased from Aldrich Chemical Co. (Milwaukee, WI). CHCA was recrystallized from methanol before using as a matrix. All other chemicals were of reagent grade. The Slide-A-Lyzer 10K dialysis cassettes (10 000 MW cutoff) were purchased from Pierce. MALDI-TOF. MALDI mass spectra were obtained on a Voyager-RP time-of-flight mass spectrometer (PerSeptive Biosystems, Framingham, MA) in the linear mode with an accelerating voltage of 30 kV and a 1.3 m flight path. A saturated solution of R-cyano-4-hydroxycinnamic acid dissolved in EtOH/H2O/HCO2H (45:45:10, v/v/v) or water/acetonitrile (2:1, v/v) was used as a matrix. The instrument is equipped with a nitrogen laser operating at 337 nm. The spectra presented in this paper represent averages of 64 or 128 laser shots. To analyze the fragment ions that dissociate from the original ion in the flight tube, post-source decay (PSD) analysis was carried out using timed ion selection, which allows only ions of a selected mass of interest to pass to the detector, in the reflector mode (24). These PSD experiments are the reflectron time-of-flight equivalents of MS/MS experiments. The time-of-flight data were externally calibrated
Yasuzawa and Tomer
using bovine serum albumin, horse heart cytochrome c, oxidized insulin B chain, angiotensin I, and CHCA as standards. HPLC. Two model 6000A chromatographic pumps, a model 441 UV detector, and a model 660 solvent programmer (Waters, Milford, MA) were used for all HPLC analyses. Separations were carried out on a Protein C-4 (250 × 4.6 mm inside diameter, 10 µm particles) reverse phase column (Vydac, Hesperia, CA) using as solvents 0.1% TFA in water (A) and in acetonitrile (B) at a flow rate of 1 mL/min. The gradients used were 5 to 70% B over 20 min for the KW-2149-HSA conjugate and 0 to 65% B over 50 min for the tryptic digests of HSA and the KW-2149-HSA conjugate. Preparation of the KW-2149-HSA Conjugate. HSA (240 mg, 3.61 mmol) and KW-2149 (10.8 mg, 18.1 mmol) were dissolved in water (2.2 mL) and incubated at 37 °C for 24 h. After incubation, the protein was dialyzed against water at 4 °C for 12 h and then lyophilized. Determination of the Binding Ratio of KW-2149 to HSA. The binding ratio of KW-2149 (mitomycin C moiety) to HSA was estimated by UV absorbance at 375 nm using KW-2149 as a calibration standard. The concentrations of the KW-2149 standard were 3, 6, 12, and 24 µg/mL, and that of the KW-2149-HSA conjugate was 2 mg/mL. Tryptic Digestion. To a solution of the KW-2149HSA conjugate (2 mg) in 50 mM ammonium bicarbonate at pH 7.9 (0.5 mL) was added trypsin (14 µg), and the solution was incubated at 37 °C for 24 h. The hydrolysis was stopped by lyophilization. The tryptic digestion of HSA was carried out in the same manner. Carboxypeptidase P Digestion. To a solution of each conjugate between T5 and KW-2149 in 50 mM ammonium acetate at pH 4.7 (100 µL) was added carboxypeptidase P (0.1 µg), and the solution was incubated at room temperature. Aliquots of the solution (20 µL) were taken at 5, 15, 30, and 60 min and 5 h and lyophilized. Each aliquot was redissolved in water (2 µL) prior to analysis by MALDI-MS. DTT Reduction of the Disulfide-Linked Peptides. To a solution of a disulfide-linked peptide in water (2 µL) was added an excess amount of 50 mM DTT (2 µL). Ambient air was displaced by nitrogen, and the sample, under nitrogen, was sealed, and allowed to stand overnight at room temperature. The reaction mixture (0.5 µL) was mixed with the matrix solution (0.5 µL) on the MALDI target and dried. RESULTS AND DISCUSSION
Preparation of the Conjugate of KW-2149 with HSA. The conjugate of HSA with KW-2149 was prepared by incubating HSA with KW-2149 (molar ratio of 1:5 HSA:KW-2149) in water at 37 °C for 24 h. Figure 3A shows the HPLC chromatogram of the incubation mixture detected at 365 nm, characteristic of the mitomycin C chromophore. There is an absorption peak, peak b, just before 20 min, which corresponds to the retention time of HSA detected at 214 nm (data not shown). This peak did not disappear after dialysis with a 10K MW cutoff (Figure 3B). This evidence clearly shows the presence of covalent binding between HSA and a species containing the mitomycin C chromophore. Peak a is unchanged KW-2149, and the remaining peaks are removed by dialysis with a 10K MW cutoff, indicating they are low-molecular weight decomposition products. The ratio of the mitomycin C moiety to HSA, determined from a KW-2149 calibration curve, was in the range of 0.5-0.8, showing that the HPLC fraction is actually a
HSA Conjugate Structure by MALDI-MS
Figure 3. HPLC chromatogram of (A) the incubation mixture of HSA with KW-2149 and (B) that after dialysis detected at 365 nm. Peak a is unchanged KW-2149, peak b is the conjugate of HSA with KW-2149, and the other peaks are the decomposition products of KW-2149.
mixture of the KW-2149-HSA conjugate and intact HSA. The structural determination of the conjugate was performed without further purification. The obtained conjugate was analyzed by MALDI-MS using CHCA as a matrix. The observed molecular weight from the mean of five measurements of HSA was 66 347.9 ( 2.6 (calcd, 66 438.2), and that of the reaction product (a mixture of conjugated and nonconjugated HSA) was 66 513.9 ( 8.6. The shift in average mass also indicates that conjugation has occurred. Peptide Mapping of HSA and the Conjugate Using MALDI Mass Spectra of an Unfractionated Tryptic Digest. As KW-2149 contains a disulfide bond, it was proposed that KW-2149 is linked to a cysteine residue(s) in HSA through a disulfide linkage which results from a disproportionation reaction between the two sulfides. To determine the position on HSA to which KW-2149 binds, tryptic digestions of intact and of KW2149-linked HSA were performed, and peptide maps were obtained. These peptide maps, especially the disulfidelinked peptides, were compared. HSA. Reduced HSA is predicted to give rise to 79 fragments by tryptic digestion (Figure 1). In nonreduced HSA, some of those peptides are linked through disulfide bonds (T9-T10-T12, T14-T21-T22, T28-T36, T38T40-T41, T42-T49, T51-T58-T61-T64-T65, and T66T75-T77). The tryptic peptides obtained by digestion of HSA were analyzed by MALDI (Figure 4A) using EtOH/H2O/HCOOH as a solvent in target preparation. The peptides observed, including six of seven expected
Bioconjugate Chem., Vol. 8, No. 3, 1997 393
disulfide-linked peptides, covered 396 amino acid residues in HSA. Using water/acetonitrile as a solvent for target preparation to obtain the peaks of small peptides, 34 more amino acids were determined (25). In total, 73% of the amino acid residues in HSA were identified by MALDI-MS of the unfractionated peptides, with 90% coverage obtained from a combination of analyses of fractionated and unfractionated peptides (Figure 1). T5, which contains the only free cysteine (Cys-34), and the largest disulfide-linked peptide, T51-T58-T61-T64T65, were not observed. (T5 was, however, observed when the HSA tryptic digest was reduced with DTT prior to analysis). For the latter peptide, however, the smaller peptides corresponding to reductive cleavage of the disulfide linkage between T61 and T64 and between T58 and T61 (T51-T58-T61 and T61-T64-T65, respectively) as well as cleavage between each component peptide (T51, T61, T64, and T65) were observed. Such cleavages at disulfide linkage were also observed for the other disulfide-linked peptides. As mentioned above, T5 was not observed in the MALDI spectrum of the tryptic digest of unmodified HSA. Because ions arising from T5 (including parts of KW2149) and T51-T58-T61-T64-T65 were observed in the tryptic digest of the conjugate (see below), the nonobservance of T5 was not due to ionization/suppression effects in the peptide mixture or to the degradation of those molecular ions. It was therefore hypothesized that T5 existed in the tryptic digest of HSA as a set of mixed sulfides. DTT reduction of the tryptic digest of HSA shifted the weights of almost all of the peptides previously observed above 3000 to lower weights (Figure 4C, T14, T36, T41, T65, and T66 and peaks around MW 2500 are enhanced), thus confirming that the original peaks resulted from disulfide-linked peptides. In the tryptic digest of HSA (Figure 4A), there are several peaks between MW 3000 and 5000 (m/z 4772, 4446, 4307, 3454, and 3306) which can be assigned to disulfide-linked peptides which are not observed in the tryptic digest of the conjugate (Figure 4B). Some of these peaks correspond to disulfide-linked peptides of T5 with other cysteine-peptides. Also, the series of T51-T58T61-T64-T65 ions observed in the conjugate digest (Figure 4B) were not observed in the spectrum of the HSA digest (Figure 4A). T5, which was also missing in the tryptic digest of HSA (Figure 4A), was clearly observed upon DTT reduction (Figure 4C), showing that this peptide existed as a disulfide-linked peptide prior to reduction. This evidence shows that, in the case of HSA, the disproportionation reaction between T5 and disulfide-linked peptides, especially T51-T58-T61T64-T65 and related peptides, occurred during the course of the tryptic digestion, resulting in rather complex mixed disulfides. These results are comparable to previous reports on the proteolytic mapping of HSA and of HSA adducts using mass spectral characterization by fast atom bombardment ionization (26-31) and by electrospray ionization (32, 33). HSA-KW-2149 Conjugate. The MALDI mass spectrum of the tryptic digest of the HSA conjugate with KW2149 differed from that of intact HSA, especially above MW 7000 (Figure 4B). Four new peaks were observed in that range (m/z 8113, 7793, 7542, and 7228) and were assigned to the disulfide-linked peptide T51-T58-T61T64-T65 and to several peptides arising from incomplete cleavages. Furthermore, two peaks, m/z 2509 and 2637, were observed in the spectrum of the conjugate which were not observed in HSA itself. These two peaks could
394 Bioconjugate Chem., Vol. 8, No. 3, 1997
Yasuzawa and Tomer
Figure 4. MALDI mass spectra of the tryptic digest of (A) HSA, (B) the KW-2149-HSA conjugate, and (C) that of HSA after DTT reduction. The dot and hyphen in the assignments indicate uncleaved peptides and disulfide bonds, respectively. The +T63, +T59, and +T59, 63 in the inset show the missed cleavages added to T51-T58-T61-T64-T65.
not be assigned to known tryptic fragments, but do correspond in weight to adducts of T5 with each sulfide
of KW-2149. Characterization of these two peaks by HPLC separation of the tryptic digest followed by MALDI-
HSA Conjugate Structure by MALDI-MS
Figure 5. HPLC chromatogram of the tryptic digest of the KW2149-HSA conjugate detected at (A) 214 nm and (B) 365 nm.
MS is detailed below. Eighty-two percent of the amino acid sequence of HSA was identified by the MALDI-MS of unfractionated tryptic peptides of the HSA conjugate. Comparison of HPLC Fractionated Peptides from HSA and the Conjugate by MALDI-MS. To compare the tryptic peptides obtained from HSA with those obtained from the KW-2149-linked HSA in more detail, HPLC separations were carried out (Figure 5). Figure 5A shows the HPLC chromatogram of the tryptic digest of KW-2149-linked HSA detected at 214 nm. All peaks were collected, and the MALDI mass spectrum of each fraction was compared with that of the corresponding fraction from the tryptic digest of HSA (HPLC data not shown). The peaks which showed absorbance at 365 nm (Figure 5B) were also collected and analyzed by MALDIMS. Although peak a is the major peak observed at 365 nm, no molecular ion corresponding to a possible KW2149 conjugate was observed, but an ion corresponding to the protonated molecular ion of T71‚72 was observed (peak a coeluted with peak 11 detected at 214 nm; see Figure 5). This peak also cochromatographed with that of the major degradation product of KW-2149 itself. The mass spectra of the other two peaks, peaks b and c, contained ions of m/z 44 019 and 67 039, respectively. The former corresponds to the peptide from residues 5-389, and the latter is unhydrolyzed KW-2149-HSA conjugate. Since there are no tryptic peptides found that absorb at 365 nm while the intact conjugate does, we postulate that the KW-2149-linked peptide was degraded to release the mitomycin C moiety during the course of tryptic digestion. The major difference in the chromatograms of HSA and of the conjugate detected at 214 nm is the presence of several peaks in the conjugate eluting between 30 and 34 min (Figure 5A), which were assigned to disulfide-linked peptides, including T51-T58-T61T64-T65, by MALDI-MS. As for the MALDI-MS of the tryptic digest of HSA, no ion corresponding to the series of T51-T58-T61-T64-T65 peptides was observed in the MALDI mass spectra of any of the HPLC fractions.
Bioconjugate Chem., Vol. 8, No. 3, 1997 395
These results are in agreement with those obtained from the unfractionated tryptic digest of HSA (Figure 4A). The disulfide-linked peptides isolated by HPLC exhibited a characteristic pattern in MALDI-MS. Patterson and Katta reported that single disulfide-linked peptides gave molecular ions corresponding to their reduced-form peptides as well as ions corresponding to the intact peptides in MALDI-MS (34). The same phenomenon was observed in the multi-disulfide-linked peptides studied here as well. Figure 6A shows the MALDI mass spectrum of the tryptic peptide T51-T58-T61-T64-T65, in which five peptides are connected to each other through four disulfide linkages. All combinations of the reduced forms of the peptide except T58 were observed, proving unambiguously the structure of this peptide. T58 (Mr ) 352) was not observed as is often the case with low-mass peptides (
