2-Amino-9H-pyrido[2,3-b]indole (AαC) Adducts and Thiol Oxidation of Serum Albumin as Potential Biomarkers of Tobacco Smoke*

Background: The reactivity of AαC, a tobacco smoke carcinogen, was investigated with DNA and albumin of human hepatocytes. Results: Hepatocytes bioactivate AαC to metabolites, which adduct to DNA and albumin. Conclusion: Cys34 and Met329 of serum albumin are targets for AαC electrophiles. Significance: AαC forms macromolecular adducts and induces oxidative stress, which may be contributing factors to liver damage and cancer risk in smokers. 2-Amino-9H-pyrido[2,3-b]indole (AαC) is a carcinogenic heterocyclic aromatic amine formed during the combustion of tobacco. AαC undergoes bioactivation to form electrophilic N-oxidized metabolites that react with DNA to form adducts, which can lead to mutations. Many genotoxicants and toxic electrophiles react with human serum albumin (albumin); however, the chemistry of reactivity of AαC with proteins has not been studied. The genotoxic metabolites, 2-hydroxyamino-9H-pyrido[2,3-b]indole (HONH-AαC), 2-nitroso-9H-pyrido[2,3-b]indole (NO-AαC), N-acetyloxy-2-amino-9H-pyrido[2,3-b]indole (N-acetoxy-AαC), and their [13C6]AαC-labeled homologues were reacted with albumin. Sites of adduction of AαC to albumin were identified by data-dependent scanning and targeted bottom-up proteomics approaches employing ion trap and Orbitrap MS. AαC-albumin adducts were formed at Cys34, Tyr140, and Tyr150 residues when albumin was reacted with HONH-AαC or NO-AαC. Sulfenamide, sulfinamide, and sulfonamide adduct formation occurred at Cys34 (AαC-Cys34). N-Acetoxy-AαC also formed an adduct at Tyr332. Albumin-AαC adducts were characterized in human plasma treated with N-oxidized metabolites of AαC and human hepatocytes exposed to AαC. High levels of N-(deoxyguanosin-8-yl)-AαC (dG-C8-AαC) DNA adducts were formed in hepatocytes. The Cys34 was the sole amino acid of albumin to form adducts with AαC. Albumin also served as an antioxidant and scavenged reactive oxygen species generated by metabolites of AαC in hepatocytes; there was a strong decrease in reduced Cys34, whereas the levels of Cys34 sulfinic acid (Cys-SO2H), Cys34-sulfonic acid (Cys-SO3H), and Met329 sulfoxide were greatly increased. Cys34 adduction products and Cys-SO2H, Cys-SO3H, and Met329 sulfoxide may be potential biomarkers to assess exposure and oxidative stress associated with AαC and other arylamine toxicants present in tobacco smoke.

Tobacco smoke is a major risk factor not only for lung cancer but also for cancer of the liver, bladder, and gastrointestinal tract (1)(2)(3)(4). The combustion of tobacco produces many genotoxicants, including polycyclic aromatic hydrocarbons, nitrosamines, aromatic amines, and heterocyclic aromatic amines (HAAs), 2 which are potential human carcinogens (5). A␣C was originally discovered as a mutagenic pyrolysis product of protein (6) and subsequently identified in cigarette smoke at levels ranging from 60 to 250 ng/cigarette (7,8). These quantities are far greater than those of the aromatic amines 4-aminobiphenyl and 2-naphthylamine, which are implicated in the pathogenesis of bladder cancer in smokers (1,9). Apart from the endocyclic nitrogen atoms, A␣C shares the same structure as 2-aminofluorene, one of the most well studied aromatic amine carcinogens (10). Significant levels of A␣C were detected in the urine of male smokers of the Shanghai cohort in China, providing evidence that tobacco smoke is a major source of A␣C exposure (11). A␣C is a liver carcinogen in mice, a transgene colon mutagen, and an inducer of colonic aberrant crypt foci, an early biomarker of colon neoplasia (12)(13)(14). Therefore, A␣C could play a role in the incidence of liver or digestive tract cancers of smokers.
A␣C undergoes metabolic activation by N-oxidation of the exocyclic amine group, by cytochrome P450 (P450) enzymes, to form 2-hydroxyamino-9H-pyrido [2,3-b]indole (HONH-A␣C) ( Fig. 1) (15,16). HONH-A␣C can undergo conjugation reactions with N-acetyltransferases or sulfotransferases to form unstable esters. These metabolites undergo heterolytic cleavage to form the proposed short lived nitrenium ion of A␣C ( Fig. 1) (17), which reacts with DNA to form covalent adducts, leading to mutations (18). The genotoxic potential of A␣C has been shown in human peripheral blood lymphocytes (19), Chinese hamster ovary cells (18), and human hepatocytes (20), where high levels of A␣C-DNA adducts are formed. However, long term stable biomarkers of A␣C must be developed for implementation in molecular epidemiology studies that seek to address a role for this chemical in human cancer risk.
DNA adducts of A␣C can be measured by sensitive liquid chromatography/mass spectrometry (LC/MS)-based methods (20). However, DNA from biopsy specimens is often unavailable, which restricts the use of this biomarker. The electrophilic N-oxidized metabolites of A␣C are also expected to react with proteins (21). The biomonitoring of protein-carcinogen adducts is an alternative approach to assess exposure to hazardous chemicals. Stable carcinogen protein adducts do not undergo repair and are expected to follow the kinetics of the lifetime of the protein (22,23). The major proteins in blood are hemoglobin (Hb) with a life span of 60 days, and human serum albumin with a half-life of 21 days. The chemistry of reactivity of Hb and albumin with various genotoxicants and toxic electrophiles has been reported (24 -26), and several protein-carcinogen adducts have been employed to assess human exposures (23,27,28).
Our goal is to develop and implement protein-based biomarkers of A␣C and other HAAs (29,30) in molecular epidemiological studies designed to assess the role of HAAs in human cancers. In this study, we have examined the reactivity of albumin with N-oxidized metabolites of A␣C. Cys 34 followed by Tyr 140 and Tyr 150 of albumin were major sites of adduction of A␣C electrophiles. The Cys 34 and Met 329 residues of albumin also served as scavengers of reactive oxygen species (ROS) generated by metabolites of A␣C, and the Cys 34 sulfinic acid (Cys-SO 2 H), sulfonic acid (Cys-SO 3 H), and Met 329 sulfoxide were recovered in high yield from human hepatocytes treated with A␣C.

Experimental Procedures
Caution-A␣C is a potential human carcinogen. A␣C and its derivatives must be handled in a well ventilated fume hood with proper use of gloves and protective clothing.
Modification of Albumin and Human Plasma with N-Oxidized Metabolites of A␣C-Mixed disulfides formed at Cys 34 of commercial albumin (34) were reduced by treatment with ␤-mercaptoethanol (35). The reduced albumin was recovered in 100 mM potassium phosphate buffer (pH 7.4) employing Amicon Ultra centrifugal filters. The reduced albumin contained 0.98 mol of Cys 34 /mol of albumin upon ␤-mercaptoethanol treatment. An equimolar solution of HONH-A␣C and HNOH-[ 13 C 6 ]A␣C or N-acetoxy-A␣C and N-acetoxy-[ 13 C 6 ]-A␣C (15 nmol, in 10 l of EtOH) or NO-A␣C (30 nmol in 10 l of EtOH) was reacted with albumin (0.6 nmol, 40 g) in 1 ml of 100 mM potassium phosphate buffer (pH 7.4) at 37°C for 18 h. N-Acetoxy-A␣C and N-acetoxy-[ 13 C 6 ]A␣C were prepared in situ by adding equimolar solution of HONH-A␣C and HONH-[ 13 C 6 ]A␣C (15 nmol) to the solution of albumin, immediately followed by the addition of 450 nmol of acetic anhydride (36), and incubated at 37°C for 1 h. The unreacted A␣C metabolites were removed by ethyl acetate extraction. Human plasma (5 l, containing ϳ200 g of albumin, 3 nmol) was diluted with 1 ml of PBS and reacted with N-oxidized A␣C derivatives as described above. Albumin from plasma was purified by affinity purification by Pierce albumin depletion kit. Other studies on A␣C-albumin adduct formation were carried out using lower amounts of N-oxidized A␣C (see below).
Human Hepatocyte Cell Culture-Human samples were obtained from the Centre de Ressources Biologiques (CRB)-Santé of Rennes. The research protocol was conducted under French legal guidelines and approved by the local institutional ethics committee. Hepatocytes were isolated by a two-step collagenase perfusion, and the parenchymal cells were seeded at a density of ϳ3 ϫ 10 6 viable cells/19.5-cm 2 Petri dish in 3 ml of Williams' medium with supplements as reported (20), except fetal calf serum was replaced with human albumin (1 g/liter) pretreated with ␤-mercaptoethanol. After 2 days, the differentiated cells were incubated with A␣C (20,33).
Albumin and DNA Adduct Formation with A␣C in Hepatocytes-Metabolism studies with A␣C (0 or 50 M in DMSO, 0.01% (v/v)) were conducted for 24 h. A solution of 1:1 A␣C and [ 13 C 6 ]A␣C (50 M) was employed for characterization of A␣C-albumin adducts, whereas the DNA adduct studies employed A␣C (50 M). The culture media containing albumin were removed after 24 h of incubation and immediately stored at Ϫ80°C. The cells were washed with PBS, and cell pellets were collected by centrifugation at 3500 ϫ g for 10 min at 4°C. Cells were stored at Ϫ80°C until further use. Cell viability was determined by 3-(4,5-dimethyltiazol-2-yl)-2,5-diphenyltetrazolium bromide test, and treatment with A␣C did not decrease cell viability (37). The medium was extracted with 3 volumes of ethyl acetate, and then the albumin was recovered with Ͼ85% purity by ethanol precipitation. For some analyses, the albumin (100 g, 1.5 nmol) was alkylated with a 100-fold molar excess of IAM (150 nmol) at 37°C for 1 h. Excess IAM was removed by Amicon Ultra centrifugal filters.
The top five precursor ions for CID-MS/MS analysis with dynamic exclusion for 180 s with three repeats of 60-s repeat duration were selected for DDA. Mass tag DDA (MS tag DDA) was employed to trigger MS/MS on peptides and amino acid adducts displaying the characteristic pattern of a 1:1 isotopic mixture of A␣C and [ 13 C 6 ]A␣C (38). The partner intensity ratio was 85-100%; m/z difference of 6 (singly charged), 3 (doubly charged), and 2 (triply charged) species. A mass list of precursor ions (Table 1) for A␣C-amino acid and A␣C-peptide adducts identified from MS tag DDA was used for targeted analysis.
The tune parameters were as follows: capillary tube temperature, 270°C; spray voltage, 2.5 kV; S-lens RF level, 68(%); insource fragmentation, 5 V; collision gas, helium; normalized collision energy, 35 eV; activation Q 0.3; HCD collision gas, argon; HCD collision energy, 20 -35. Full scan data were acquired over a mass range of 150 -1800 m/z at a resolving power of 120,000 (at 400 m/z), whereas MS/MS data acquisition was carried out with the ion trap or Orbitrap (resolving power 60,000) as mass analyzer with isolation widths 1.5 and 2 m/z. An internal lock mass at m/z 371.1012 (polysiloxane) was employed while using the Orbitrap mass analyzer. The tune parameters were optimized with LQQCPFEHVK and LQQCPF peptides (New England Peptide, Gardner, MA). Mass spectral data were acquired with Xcalibur version 3.0.63 software.
Data Analysis of A␣C-Albumin-The BumberDash platform (version 1.4.115) with the Myrimatch search algorithm (version 2.1.138) with a 31-protein subset of the RefSeq human protein database, version 37.3, with their forward and reverse protein sequences were used to search for the A␣C-peptide adducts (29,39). The dynamic modifications used in the Myrimatch configuration file were as follows: oxidation (ϩ15.99 Da) on Met, Cys; deamidation (Ϫ17.03 Da) on N-terminal Gln; A␣C and [ 13 C 6 ]A␣C (ϩ181.06 and 187.11) on Cys, Lys, Tyr, Ser, Thr, Trp, His, Glu, Asp; SO-A␣C and SO-[ 13 C 6 ]A␣C (ϩ197.06, ϩ203.10) on Cys and SO 2 -A␣C and SO 2 -[ 13 C 6 ]A␣C (ϩ213.05, ϩ219.10); acetylation (ϩ42.01) on Lys, Cys, Tyr, Arg. The search parameters used were as follows: enzyme (trypsin/chymotrypsin), up to two missed cleavages, mass tolerances 1.25 m/z (precursor ions) and 0.5 m/z (product ions). Each identified spectrum was filtered by the IDPicker algorithm (version 3.0.564) with a false discovery rate of 1%. The peptide sequences of identified A␣C-peptide and A␣C-amino acid adducts were confirmed by manually sequencing.
DNA Isolation and UPLC/MS 3 Measurement of A␣C-DNA Adducts-DNA was isolated by the phenol/chloroform method (20). DNA (5 g) was spiked with the isotopically labeled internal standard ([ 13 C 10 ]dG-C8-A␣C) at 1 adduct per 10 6 bases, followed by enzymatic digestion of DNA (40). The UPLC/MS 3 measurements of A␣C-DNA adducts were performed using a NanoAcquity UPLC system (Waters Corp.) coupled with an LTQ VelosPro ion trap mass spectrometer (Fisher) as described previously (41). The ions were monitored with the MS 3

Results
The metabolic activation of A␣C in hepatocytes, the reaction of N-oxidized A␣C metabolites with DNA and albumin, and the scavenging of ROS by albumin are summarized in Fig. 1.
The UV spectra of A␣C, HONH-A␣C, and NO-A␣C are displayed in Fig (Fig. 3). Nine A␣C-peptide adducts (P1-P9) were detected in the tryptic/chymotryptic digest. The assignments of the peptide adduct sequences and accurate mass measurements of the amino acid adducts, following digestion with Pronase E, leucine aminopeptidase, and prolidase, are listed in Table 1. Cys 34 , followed by Tyr 140 and Tyr 150 , formed adducts with HONH-A␣C; an additional adduct was formed at Tyr 332 with N-acetoxy-A␣C. The adduction of NO-A␣C primarily occurred at Cys 34 , followed by much lower levels of adduct formation at Tyr 150 .
MS Characterization of Adducts Formed at Cys 34 of Albumin-The digestion of A␣C-modified albumin with trypsin/chymotrypsin produced peptides containing adducts at 31  A total of six adducts with different oxidation states of the sulfur atom were identified. The product ion spectra of several adducts peptides are described in detail below.  at m/z 514.2, a mass 16 Da less than the sulfonamide adduct. The product ion spectrum of m/z 514.2 contains minor fragment ions at m/z 183.1 (A␣C ⅐ ϩ ) and 184.1 (A␣C ϩ ), with a base peak at m/z 679.9, corresponding to the sulfonium ion [M ϩ 2H-A␣C] 2ϩ (Fig. 4B). The MS 3 scan stage product ion spectrum of the proposed LQQ*C [SO] PFEDHVK sulfoxide ion at m/z 679. 9 [MЈ] showed -b and -y ion series and supports the sequence assignment. Doubly charged product ion at m/z 648.9 attributed to the loss of CH 2 SO from the MЈ ion and *b 4 -SO, *b 5 ,*y 8 , and *y 9 ions supports the proposed structure (Fig. 4C).
LQQ*CPF-A␣C Peptide Adducts-Three additional S-Nlinked peptide adducts were identified. The product ion mass spectra of the doubly protonated ions are consistent with the A␣C sulfonamide LQQ*C  (41,42). The prominent base peak observed at m/z 184.1 is attributed to protonated A␣C (Fig. 4G). It is noteworthy that most arylsulfenamide adducts undergo hydrolysis during proteolytic digestion, which generally makes these adducts difficult to detect (30,43,44).
MS Characterization of *Cys-A␣C Adducts-The Cys sulfenamide-and sulfinamide-linked adducts of A␣C underwent hydrolysis to produce A␣C during proteolysis of albumin with Pronase E, leucine aminopeptidase, and prolidase. Two peaks attributed to Cys-A␣C adducts containing the S-dioxide linkage were identified (t R ϭ 10.9 and 11.5 min) (A1) in the UPLC/MS chromatogram (Fig. 5, A and B). The precursor ions [M ϩ H] ϩ were observed at m/z 335.0806, a value that is within 0.6 ppm of the calculated m/z value of the proposed sulfonamide structure ( Table 1). The adducts were present in an approximate ratio of 4:6 in albumin modified with HONH-A␣C and in a 2:8 ratio when albumin was modified with N-acetoxy-A␣C; Fig. 5, A and B). The adducts were formed at ϳ5-fold higher levels in albumin treated with N-acetoxy-A␣C than in albumin treated with HONH-A␣C. The earlier eluting isomer underwent CID to preferentially form a radical cation at m/z 183.0790 (A␣C ⅐ ϩ ), whereas the second adduct favored the formation of the even electron ion at m/z 184.0867, the m/z of protonated ion of A␣C ϩ (Fig. 5 (C and D) and Table 2). Other  (Fig. 5 (C and D) and Table 2).
Arylsulfonamides are weak acids, and the nitrogen anion of the sulfonamide linkage can form several tautomeric forms with hindered rotation about the SϭN bond (45). However, the  prominent differences in the product ion spectra, combined with the inability to interconvert these Cys-A␣C adducts at elevated temperature, suggest that the adducts are not conformational isomers.
The nitrenium ion of HONH-A␣C can undergo charge delocalization to form the carbenium ion resonance form with electron deficiency centered at the C-3 position of the A␣C skeleton ( Fig. 5E) (46). We propose that one isomeric Cys-A␣C adduct contains an S-N linkage, and the second adduct contains a thioether linkage, formed between the Cys 34 -SH group and possibly the C-3 atom of A␣C (46). The S-N-sulfenamide (or sulfinamide) and thioether adducts undergo oxidation, by ROS generated by aerobic oxidation of N-oxidized A␣C metabolites (Fig. 1) or during proteolysis, to form the sulfonamide and sulfone linkages (Fig. 5E) (47). Both adducts undergo CID to form the proposed sulfonium ion at m/z 230.0385 (Fig. 5E). These adducts were not chromatographically resolved in the tryptic/chymotryptic digests of albumin. Large scale syntheses and NMR studies are required to elucidate the respective structures of these A␣C-Cys-S-dioxide linked isomers. The Cys 34 of albumin was reported to react with N-acetyl-p-benzoquinoneimine, the electrophilic metabolite of acetaminophen, and formed two regioisomeric adducts (42).
MS Characterization of Adducts Formed at Tyr 140 , Tyr 150 , and Tyr 332 of Albumin-Three adducts were formed between Tyr residues of albumin and A␣C (Table 1).  (Fig. 6, A and B). We recently reported a similar mechanism of fragmentation of an O-linked adduct formed between tyrosine and the HAA 2-amino-1-methyl-6-phenylmidazo[4,5-b]pyridine (PhIP) (32).     JUNE  are proposed to arise by cleavage of the C ␣ and C ␤ bond of tyrosine. The fragment ion at m/z 183.0791 is assigned to A␣C ⅐ ϩ , and the ion at m/z 198.0659 is tentatively assigned as the protonated ion of the 2-imino-3-oxo derivative of A␣C, with the loss of phenylalanine (165.0790 Da) as a neutral fragment. The ion at m/z 198.0659 provides evidence that adduct formation occurred between the 4-HO group of tyrosine and the A␣C heterocyclic ring (Fig. 6D).

Characterization of A␣C-Serum Albumin Adducts
F*YAPELL-A␣C Adducts-Two minor Tyr adducts of A␣C with the peptide sequence 149 F*Y [A␣C] APELL 155 (P9A and P9B; Fig. 3 and Table 1) were identified at t R 14.5 and 15.8 min (Fig.  7A). The full scan spectra of both adducts displayed doubly charged [M ϩ 2H] 2ϩ at m/z 517.3. The CID fragment ions of m/z 517.3 displaced the same characteristic -b ion and -y ion series attributed to the F*Y [A␣C] APELL for both peptide adducts (Fig. 7A, CID). The shift in mass between the y 5 and *y 6 ion series proves that the site of A␣C adduction occurred at Tyr 150 for both peptides (Fig. 7A). CID did not provide appreciable fragment ions at the low m/z region for either adduct. The doubly protonated ions [M ϩ 2H] 2ϩ at m/z 517.3 were subjected to HCD to examine for potential fragmentation of the bond formed between the Tyr and A␣C. The extracted ion chromatograms were generated at m/z 183.0791 and m/z 184.0869 (Fig. 7A, HCD). The earlier eluting adduct (P9A) at t R ϭ 14.5 min displayed a prominent fragment ion at m/z 184.0869, an ion attributed to protonated A␣C (Fig. 7A, HCD at t R ϭ 14.5). The linkage of this adduct may have occurred between the C-3 or C-5 atom of the Tyr phenyl ring and the exocyclic amino group of A␣C. The HCD product ion spectrum of the second adduct (P9C) at t R ϭ 15.8 min showed a mixture of product ions assigned as the radical cation of A␣C ⅐ ϩ and protonated A␣C, respectively, at m/z 183.0789 and m/z 184.0868 (Fig. 7A, HCD at t R ϭ 15.8). A Tyr linkage may have formed at the C-3 or C-7 atom of the heterocyclic ring of A␣C, based on studies of nucleophilic trapping agents adducting at these sites of the nitrenium ion of A␣C (46,48).

MS Characterization of an Amine-linked A␣C Tyr Adduct-
The second set of Tyr-A␣C and Tyr-[ 13 C 6 ]A␣C adducts (A3) eluted at t R ϭ 12.1 min (Fig. 7B). The protonated ions [M ϩ H] ϩ were observed at m/z 361.1293 and 367.1475, respectively, a mass 2 Da less than the Tyr-A␣C and Tyr-[ 13 C 6 ]-A␣C adducts (m/z 363.1449 and 369.1631) described above (Fig. 7B and Table 2). The structure of the adduct is proposed to be a quinoneimine-linked adduct, which occurs by oxidation of P9A (Fig. 7B) during proteolytic digestion with the three-enzyme mixture. The product ion spectrum of [M ϩ H] ϩ at m/z 361.1293 shows ions at m/z 344.1025, 315.1239, and 300.1130, which are of losses of NH 3 , H 2 CO 2 , and NH 3 and CO 2 , respec- tively ( Fig. 7B and Table 2). The product ion at m/z 288.1129 is proposed to arise following ring opening of the protonated quinoneimine ring (Fig. 7C). The Tyr-[ 13 C 6 ]A␣C adduct displays the same product ions as the unlabeled adduct but shifted by 6 m/z.
A third peptide adduct formed between A␣C and Tyr was observed only in the albumin modified with N-acetoxy-A␣C. The CID-MS/MS spectrum of doubly charged protonated precursor ion at m/z 616.8 at 19.3 min displayed a typical -b and -y ion pattern attributable to LGMFL*Y [A␣C] EY (P6) with A␣C adduction at Tyr 332 (supplemental Fig. S1, top). The [ 13 C 6 ]A␣C homologue of this adduct at m/z 619.8 displayed the same pattern of fragmentation (supplemental Fig. S1, bottom).

A␣C-SA Adduct Formation as a Function of Concentration of N-Oxidized A␣C Metabolites-
The LQQ*C [SOA␣C] PFEDHVK (P3) sulfinamide adduct at Cys 34 accounted for 73-80% of the total ion counts of A␣C-peptide adducts, when commercial albumin was reacted with a 50-fold molar excess of N-oxidized A␣C metabolites. In vivo, the exposure to A␣C occurs at much lower levels than the physiological concentration of albumin, and relative abundances of adducts formed may be different from those adducts formed at elevated exposures to A␣C in vitro. Therefore, albumin was treated with HONH-A␣C or N-acetoxy-A␣C over a million-fold range of carcinogen/mol of albumin (1-10 Ϫ6 ) (Fig. 8, A and B). We also examined the effect of plasma matrix components on the reactivity of albumin with N-oxidized A␣C metabolites. Representative UPLC/MS chro-matograms of the peptide adducts recovered from commercial albumin or albumin in plasma modified with 1 molar eq of N-acetoxy-A␣C are shown in supplemental Fig. S2, A and B.
The Tyr-peptide adducts of A␣C at Tyr 140 , Tyr 150 , and Tyr 332 residues were only detected when albumin was reacted with Ն 0.01 mol N-acetoxy-A␣C per mol albumin. In contrast, adducts were still formed at Cys 34 with a 1 ϫ 10 Ϫ5 molar ratio of N-acetoxy-A␣C/albumin (Fig. 8, A and B), and LQQ*C [SOA␣C] PFEDHVK sulfinamide was the major adduct. The amounts of A␣C-peptide adducts formed were ϳ5-40 times higher in reactions of albumin conducted with N-acetoxy-A␣C than those amounts of albumin adducts formed with HONH-A␣C (Fig. 8, A and B). The level of A␣C adduct formation with albumin in plasma was severalfold lower than adduct levels formed with commercial albumin (data not shown). Plasma albumin adducts were not detected at Tyr 140 , Tyr 150 , or Tyr 332 residues at any concentration of N-oxidized A␣C, and the sulfinamide LQQ*C [SOA␣C] PFEDHVK was the predominant adduct.
A␣C-DNA and A␣C-Albumin Adduct Formation in Human Hepatocytes-Reactive N-oxidized intermediates of A␣C are formed and adduct to DNA in human hepatocytes and albumin in human hepatocytes (Fig. 1) (20, 33). The basal activities of P450 1A1 and 1A2, two major isoforms involved in N-oxidation of A␣C (16,17), were measured in human hepatocytes from three donors using ethoxyresorufin and methoxyresorufin as substrates (33). The level of dG-C8-A␣C adduct formation

Characterization of A␣C-Serum Albumin Adducts
from the two donors (A and C) with the highest P450 1A1 and 1A2 activities produced higher levels of dG-C8-A␣C than donor B with low enzyme activity (Fig. 9A).
The mass chromatograms of A␣C-albumin peptide adducts, following tryptic/chymotryptic digestion of albumin recovered from hepatocytes, are shown in Fig. 9B. The relative ion abundances of the peptide adducts are summarized in Fig. 9C, and product ion spectra are presented in supplemental Fig. S3. The major adduct, based on ion counts, is LQQ*C [SOA␣C] -PFEDHVK sulfinamide, followed by LQQ*C [SOA␣C] PF and LQQ*C [A␣C] PF. The occurrence of the LQQ*C [SOA␣C] -PFEDHVK as the major adduct in hepatocytes is similar to the findings of commercial albumin and albumin in plasma treated with N-oxidized A␣C metabolites (Figs. 7A, 8A, and 9 (B and C)). The LQQ*C [A␣C] PF sulfenamide and LQQ*C [SOA␣C] PF sulfinamide were also identified but occurred at lower ion abundances ( Fig. 9 (B and C) and supplemental Fig. S3 (A-C)). LQQ*C [SO2A␣C] PF and LQQ*C [SO2A␣C] PFEDHVK were not detected.
In the absence of stable, isotopically labeled internal standards, quantitative peptide adduct measurements and correlations to DNA adduct levels of A␣C cannot be determined. However, the ion counts of the major albumin peptide adduct, LQQ*C [SOA␣C] PFEDHVK sulfinamide, were greatest in donor C, who also harbored the highest level of dG-C8-A␣C (Fig. 9, A-C). Very high levels of A␣C also were recovered in all of the hepatocytes (Fig. 9, B and C).The occurrence of A␣C is attributed to hydrolysis of S-N-linked albumin-A␣C adducts during proteolysis and not to unmetabolized A␣C bound to albumin, because the isolation procedure effectively removed all unbound A␣C from albumin in the cell culture media. 3 Oxidative Status of Albumin-Cys 34 and Albumin-Met 329 in Human Hepatocytes Exposed to A␣C-N-Oxidized metabolites of arylamines generate ROS (49). We sought to determine if A␣C had induced oxidative stress in hepatocytes by identification of oxidation products of Cys 34 and Met residues of albumin (50 -52). The oxidized sulfinic LQQ*C [SO2H] PF and sulfonic LQQ*C [SO3H] PFEDHVK acids of Cys 34 of albumin were monitored (29,32), and LQQCPF was measured following derivatization of albumin with IAM (29). Elevated levels of the protonated [M ϩ H] 2ϩ peptides for LQQ*C [SO2H] PFEDHVK at m/z 688.3 and LQQ*C [SO3H] PFEDHVK at m/z 696.3 (29) were detected in all three sets of hepatocytes treated with A␣C (Fig.  9D). The product ion spectra at m/z 688.3 and 696.3 displayed typical -b and -y ion series type fragment ions (supplemental Fig. S4, A and B) and permitted the assignments as the peptide sequences of the Cys sulfinic and sulfonic acids (29). The levels of LQQ*C [SO2H] PFEDHVK and LQQ*C [SO3H] PFEDHVK in A␣C/[ 13 C 6 ]A␣C-treated hepatocytes were 6 -8 and 4 -6 times higher, respectively, than the levels in untreated controls, and the levels of IAM-derivatized LQQ*CPF were decreased by more than 10-fold in A␣C-treated hepatocytes (Fig. 9, B and D).  Fig. S4C). From targeted analysis, the level of DVFLGM [O] F was ϳ6 times higher in A␣C/[ 13 C 6 ]A␣C-treated hepatocytes than in untreated hepatocytes (Fig. 9, B and D).

Discussion
Primary human hepatocytes are an ideal ex vivo model system for studying metabolism, bioactivation, and mechanisms of toxicity of carcinogens, because cofactors are present at physiological concentrations, and biotransformation pathways may closely simulate those that occur in vivo (53). In this study, we investigated the metabolic activation of A␣C, a rodent liver carcinogen, in human hepatocytes and examined the reactivity of its genotoxic N-oxidized metabolites of A␣C with DNA and 3 K. Phatak, unpublished observations. albumin. Our goal is to develop and implement albumin-based biomarkers of A␣C and other HAAs in molecular epidemiological studies designed to assess the role of HAAs in human cancers (54). The Cys 34 residue was the major nucleophilic site of albumin to form adducts with A␣C, followed by Tyr 140 and Tyr 150 , which formed adducts at minor levels. Another HAA, PhIP (55), also primarily formed adducts at the Cys 34 of albumin with considerably lower levels of adducts occurring at Tyr 140 and Tyr 150 (56). In contrast, the Tyr 140 and Tyr 150 residues of albumin are the preferred site for adduct formation of neurotoxic organophosphate compounds (57)(58)(59).
The preliminary characterization A␣C-albumin adducts was performed in vitro. The formation of A␣C-albumin adducts was greatly enhanced by the in situ generation of N-acetoxy-A␣C, a reactive intermediate of HONH-A␣C, which undergoes heterolytic cleavage to produce the nitrenium ion (48). N-Acetoxy-A␣C is a penultimate metabolite of A␣C that adducts to DNA and proteins (17). The N-acetoxy intermediates of HON-A␣C and other N-hydroxylated HAAs, such as 2-hydroxyamino-3-methylimidazo[4,5-f]quinoline and 2-hydroxyamino-3,8-dimethylimidazo [4,5-f]quinoxaline, are unstable and cannot be isolated (60). However, the in situ formation of these N-acetoxy intermediates in the presence of DNA, by reaction of the N-hydroxylated HAAs with acetic anhydride, increased the levels of DNA adducts by 10 -30-fold (36,61,62). Using a similar reaction scheme, we showed that adduct formation of HON-A␣C with albumin was also greatly enhanced by the in situ generation of N-acetoxy-A␣C (Fig. 8, A and B). The Lys, Arg, Cys, and Tyr residues of albumin can potentially compete with HONH-A␣C and undergo acetylation with acetic anhydride and influence the formation of different A␣C-albumin adducts. However, the same A␣C-albumin adducts were formed by reaction of albumin with a 50-fold molar excess of HONH-A␣C, NO-A␣C, or N-acetoxy-A␣C generated in situ (Table 1). Moreover, the Myrimatch search engine did not detect acetylation at Lys, Arg, Cys, or Tyr residues of albumin under these reaction conditions (100 mM potassium phosphate buffer (pH 7.4, 37°C). Under low reaction conditions with HONH-A␣C (10,000:1, albumin/HONH-A␣C), the levels of Cys 34 adducts were 50-fold greater when albumin was reacted with HONH- A␣C in the presence of acetic anhydride than by reaction of albumin with HONH-A␣C alone, thereby demonstrating that N-acetoxy-A␣C is efficiently formed in situ and readily reacts with nucleophilic sites of albumin to form covalent adducts (Fig. 8, A and B).
Human hepatocytes efficiently bioactivated A␣C to electrophilic N-oxidized metabolites, which formed covalent adducts with DNA and albumin (Fig. 9). The DNA adduct, dG-C8-A␣C, was formed at relatively high levels, ranging from 2 to 12 adducts/10 6 DNA bases, consistent with our previous data (20,33). dG-C8-A␣C was also previously identified in salivary DNA of smokers (63). The Cys 34 residue was the sole site of albumin found to form adducts with A␣C metabolites in hepatocytes; both sulfenamide and sulfinamide adducts were identified. However, the ion counts of A␣C recovered from the albumin digest were 100-fold or greater than the ion counts of any of the A␣C-Cys adducts. These findings signify that a large proportion of the S-N-linked albumin-A␣C adducts underwent hydrolysis during proteolysis. In the absence of stable peptide adducts or isotopically labeled LQQ*C [SOA␣C] PFEDHVK peptides for internal standards, it is difficult to determine the relative reactivity of N-oxidized A␣C metabolites with DNA and albumin in hepatocytes.
HONH-A␣C and NO-A␣C undergo redox cycling in hepatocytes and produce ROS (Fig. 1) (49, 64). Previous studies reported that structurally related aromatic amines, some of which are present in tobacco smoke (1,66), deplete glutathione levels in the liver or ex vivo in hepatocytes of rodents and induce oxidative DNA damage (67)(68)(69); however, the oxidation of albumin was not reported in those studies. In our study, we show that albumin scavenges ROS produced by metabolites of A␣C in human hepatocytes by formation of the oxidized Cys 34 -containing peptides LQQ*C [SO2H] -PFEDHVK, LQQ*C [SO3H] PFEDHVK, and the methionine oxidation peptide DVFLGM [O] F. The level of Cys 34 of albumin alkylated with IAM prior to proteolytic digestion decreased by more than 90% in hepatocytes treated with A␣C and provides strong evidence that Cys 34 of albumin is scavenging ROS (Fig. 9C). Together with Cys 34 , six Met residues of albumin display antioxidant activity toward O 2 . , H 2 O 2 , and HOCl (51,52). Met 329 was the primary site of Met oxidation in albumin from human hepatocytes treated with A␣C. The oxidation of Met residues, particularly Met 329 , has been observed in albumin of hemodialysis patients (52). The major adducts of A␣C formed with albumin were determined in human hepatocytes by employing data-dependent scanning and bottom-up proteomics approaches, and the adduction products were the same as those adducts formed in vitro with commercial albumin reacted with N-oxidized A␣C intermediates. The dose of A␣C (50 M) employed in human hepatocytes is greater than daily human exposure to A␣C but comparable with the doses employed in studies investigating the genotoxicity of A␣C (18,19). The amount of A␣C arising in mainstream smoke is 60 -250 ng/cigarette (7,8). Thus, the intracellular or plasma levels of A␣C found in humans exposed to this tobacco carcinogen are considerably lower than the amount of A␣C employed in our hepatocyte study. However, dG-C8-A␣C formation occurs in a concentration-dependent manner in human hepatocytes treated with A␣C over a 10,000fold concentration range (1 nM to 10 M), signifying that the reactive N-oxidation metabolites of A␣C are formed at physiological exposure levels (20). More sensitive mass spectrometrybased methods are required for measuring A␣C-albumin adducts and albumin oxidation products in hepatocytes at these lower exposure conditions to A␣C.
The Cys 34 of albumin accounts for 80% of total free thiol content in plasma and is considered a major antioxidant and scavenger of electrophiles in plasma (34). A number of genotoxicants and toxic electrophiles form adducts with the Cys 34 of rodent or human albumin (23). Many of these adducts have been characterized primarily in vitro, and several adducts have been detected in humans. Albumin adducts have been identified with acrylamide (70), nitrogen mustard (71), ␣,␤-unsaturated aldehydes (26), the neurotoxin brevetoxin B (72), acetaminophen (42), benzene (23), and several N-oxidized HAAs of diverse structures (32,35,56,73). In addition, aldehydes produced in tobacco smoke are scavenged by Cys 34 of albumin in vitro (74). Immunochemical techniques have shown carbonyl residues are formed with albumin exposed to cigarette smoke extract, and the carbonylation of albumin, by tobacco smoke, has been detected in lung tissue biopsy samples of smokers (75). Spectrophotometric assays have shown a decrease in the free Cys 34 content of albumin following exposure to cigarette smoke extracts in vitro (74,76). The diminution of free Cys 34 content may be attributed to adducts formed with aldehydes or by oxidation with ROS (26,74,77); however, a correlation between the level of adduct formation at Cys 34 of albumin and cigarette smoking constituents remains to established in vivo.
Recently, elevated levels of Cys 34 -SO 2 H of albumin were detected in plasma of smokers in a small pilot study (65). A␣C, other HAAs, and structurally related aromatic amines present in tobacco smoke represent a class of chemicals in tobacco smoke that may contribute to the chemical modification or oxidation of Cys 34 and the oxidation of Met residues of albumin.
In summary, A␣C, a rodent liver carcinogen (12), undergoes bioactivation and forms adducts with DNA and albumin and induces oxidative stress in human hepatocytes. Albumin is a potent scavenger of ROS species generated by A␣C metabolites. A␣C, other HAAs, and many aromatic amines that arise in mainstream tobacco smoke (1,5,66) undergo N-oxidation in humans (54). Some of these metabolites form adducts with DNA and protein and also induce oxidative stress, which may be contributing factors to liver damage and cancer risk in smokers. The Cys 34 adducts of N-oxidized HAAs and arylamines or their hydrolysis products and elevated levels of Cys 34 -SO 2 H and Cys 34 -SO 3 H of albumin may be potential biomarkers to assess exposure to these hazardous chemicals in tobacco smokers.