Organ-Specific Screening for Protein Damage Using Magnetic Bead Bioreactors and LC–MS/MS

Abstract

A new 96-well plate methodology for fast, enzyme-multiplexed screening for metabolite-protein adducts was developed. Magnetic beads coated with metabolic enzymes were used to make potentially reactive metabolites that can react with test protein in the wells, followed by sample workup in multiple 96-well filter plates for LC-MS/MS analysis. Incorporation of human microsomes from multiple organs and selected supersomes of single cytochrome P450 (cyt P450) enzymes on the magnetic beads provided a broad spectrum of metabolic enzymes. The reacted protein was then isolated, denatured, reduced, alkylated, and digested, and peptides were collected in a sequence of 96-well filter plates for analysis. Method performance was evaluated by trapping acetaminophen reactive metabolite N-acetyl-p-benzoquinoneimine (NAPQI) with human glutathione S-transferase pi (hGSTP), human serum albumin (HSA), and bovine serum albumin (BSA) as model target proteins. Relative amounts of acetaminophen metabolite and hGSTP adducts were compared with 10 different cyt P450 enzymes. Human liver microsomes and CYP1A2 supersomes showed the highest bioactivation rate for adduct formation, in which all four cysteines of hGSTP reacted with NAPQI. Eight cysteines of HSA and four cysteines of BSA have been detected to react with NAPQI. This method has the potential for fast multienzyme protein adduct screening with high efficiency and accuracy.

Graphical Abstract

Protein damage by drugs or pollutants most often involves bioactivation to a reactive electrophilic metabolite followed by reaction with a nucleophilic moiety on proteins.^1–3 These metabolite reactions on proteins can change or destroy protein functions or provoke an antigen response, resulting in organ toxicity in humans.^4,5 Drug-protein adducts are often monitored in the pharmaceutical industry by using radioisotope tracer in vitro assays.⁶ Other methods include the enzyme-linked immunosorbent assay and Western blots and liquid chromatography-mass spectrometry^7,8 that have not implemented multiplexed enzyme or magnetic bead approaches.

Liquid chromatography-mass spectrometry (LC-MS) proteomic analysis is a powerful tool to detect protein modifications.^9–11 Typical shotgun proteomics starts with digestion of proteins in the mixture and separation of resulting peptides by HPLC and then identifies the peptides by tandem MS. A reliable, highly parallel sample preparation methodology to facilitate multiple-enzyme protein damage screening would be a valuable tool in molecular toxicity assessment and improve cost-effectiveness.^12–14

96-well filter plates (96FASP) are commonly used in protein sample preparation methods^14,15 to accomplish buffer exchange, denaturation, reduction, alkylation, and digestion. The weakness of sample preparation in this format is low centrifugation speed, limited to ~2000g. Therefore, many washing steps with centrifugation take 4–6 h before digestion. Protein precipitation with acetone under a low centrifuge speed in 96-well plates (96PACS) can be used as a faster high-throughput sample preparation approach.¹⁶ Arul et al. developed automated, high-throughput sample preparation for proteins in a 96-well filter plate by using vacuum pressure without centrifugation or transfer of the filter plate (4–5 h before digestion).^12,17 In this paper, we modified this approach for metabolite-protein reaction studies by first using enzyme-coated magnetic beads to generate metabolites from a test chemical and then achieve buffer exchange, denaturation, reduction, and alkylation on a filter plate under vacuum to decrease the preparation time to ~2 h before digestion (Scheme 1).

Scheme 1. Multiplexed Enzyme Reactions Followed by Sample Workup and LC-MS/MS Sequencing to Determine Protein Reaction Sites and Quantify Relative Product Amounts^a.

^a(A) Magnetic beads (MB) coated with cyt P450 microsomes and supersomes coupled to NADPH regeneration in a 96-well plate convert APAP to NAPQI that reacts with hGSTP. (B) Solutions of modified hGSTP were separated from MBs by transfer to a 96-well filter plate. (C) Modified hGSTP was reduced and alkylated. (D) Modified hGSTP was digested, and peptides were filtered down into a new plate. (E) LC-MS/MS analysis. Center panel shows a 96-well plate with an experimental reaction plan.

Acetaminophen (APAP), a common analgesic,¹⁸ was chosen to investigate drug-induced protein damage, a leading cause of acute liver failure in APAP users.¹⁹ After being absorbed by the gastrointestinal tract, most acetaminophen is converted to nontoxic glucuronide (acetaminophen-gluc) and sulfate (acetaminophen-sulf) metabolites by UDP-glucuronosyltransferases and sulfotransferases. However, 5–9% of acetaminophen is oxidized to N-acetyl-p-benzoquinoneimine (NAPQI) catalyzed by cytochrome P450 (cyt P450) enzymes.^20,21 Usually, NAPQI is rapidly detoxified by glutathione (GSH), but in situations of GSH deficiency, such as acetaminophen overdose or long-term use, excess NAPQI reacts with cysteine residues in proteins, causing cell death and toxicity in the liver.^22,23 Human glutathione S-transferase pi (hGSTP), the most effective enzyme catalyzing GSH conjugation with electrophiles, is one of the binding target proteins for excess NAPQI.^24–26 Albumins are the most abundant proteins in blood serum. Many carcinogens or drugs and their metabolites were reported to react with albumins to form adducts.²⁷ Here, HSA and BSA are chosen as target proteins in our experiments. Supersomes representing the most abundant cyt P450 enzymes in each organ were selected to test the system, i.e., CYPs 1A2 (liver), 2A6 (lung), 1B1 (kidney), and 3A5 (intestine). Also, CYP 3A4 and 2E1, reported as major cyt p450 enzymes for acetaminophen metabolism were included, as were human liver (HLM), human lung (HLuM), kidney (HKM), and intestine (HIM) microsomes.

We previously identified and detected hGSTP-NAPQI adducts by using magnetic beads (MBs) coated with cyt P450 2E1 as bioreactors.28 In the present paper, we extend this approach to multiple-organ microsomal and supersomal enzyme sources to evaluate relative amounts of NAPQI-GSH conjugate. The chemically modified positions of NAPQI on hGSTP, HSA, and BSA were identified, and relative amounts of reacted peptides were compared using a 96-well plate sample workup format developed for LC-MS/MS analysis (Scheme 1). Raloxifene was also used as a test compound to evaluate our approach. All four cysteines (Cys-14, Cys-47, Cys-101, Cys-169) in hGSTP, eight cysteines (Cys-98, Cys-124, Cys-200, Cys-223, Cys-288, Cys-339, Cys-499, Cys-500) in HSA, and four cysteines (Cys-199, Cys-222, Cys-287, Cys-338) in BSA were found to react with NAPQI and form 3-(cysteine-S-yl)-acetaminophen adducts. Liver and lung microsomes and the CYP1A2 supersome were the most effective enzyme sources for bioactivation of acetaminophen and damage to hGSTP.

EXPERIMENTAL SECTION

Chemicals and Reagents.

Carboxylated magnetic beads (MB, 1 μm diam., 20 mg/mL) were from Polysciences, Inc. Glutathione S-transferase (GST) from human placenta, albumin from human serum, and bovine serum albumin were from Sigma-Aldrich. Pierce trypsin protease, Pierce Glu-C protease, and Pierce chymotrypsin protease (TLCK treated) were from ThermoFisher Scientific. Sources of all chemicals and full experimental details are in the Supporting Information (SI) file.

Sample Workup.

Layer-by-layer (LbL) assembly of films of microsomes and cytochrome p450 supersomes on magnetic beads (MB/PDDA/PSS/PDDA/enzymes) were similar to those described previously.^28–31 Full details are described in the SI file.

Metabolite Formation.

The magnetic bioreactors were incubated with 1 mM acetaminophen, 1 mM GSH, and an NADPH regeneration system (10 mM G6P, 1 mM NADP⁺, 1 U/mL G6PDH enzyme, 1 mM MgCl₂) in 200 μL of Tris buffer. After incubation at 37 °C for 1 h, the reaction was terminated by magnetic separation. Solutions were transferred using pipettes to a 96-well Omega membrane 3K MWCO filter plate, filtered under vacuum into a 96-well plate, and spiked with 0.14 μM of N²-benzoylguanosine as an internal standard before LC-MS/MS analysis. All reactions were done in triplicate.

Protein Adduct Formation.

hGSTP (or HSA, BSA) (100 μg) was incubated with 20 mM DTT at 37 °C for 1 h, washed twice and concentrated by a 3 kDa Ultra centrifugal filter unit, and resuspended in 100 μL of Tris buffer. The magnetic bioreactors were incubated with 1 mM acetaminophen or raloxifene, 100 μg of pretreated protein, 2 mM dithiothreitol (DTT), and an NADPH regeneration system (10 mM G6P, 1 mM NADP⁺, 1 U/mL G6PDH enzyme, 1 mM MgCl₂) in 200 μL of Tris buffer. After incubation at 37 °C for 12 h, the reaction was terminated by magnetic separation. The solutions were transferred to 96-well Omega membrane 3K MWCO filter plate. The control experiment was done in the same conditions without adding acetaminophen. All reactions were done in triplicate.

Sample Preparation.

All sample treatments were done during vacuum filtrations (Scheme 2). Protein denaturation, disulfide group reduction, and thiol group alkylation were achieved while solutions passed through the filter membrane. The vacuum was 15 inHg, and the solution took 30 min to pass through the filter membranes. A 96-well plate cover was placed under the plate filter to collect waste. First, samples were washed with 100 μL of 50 mM ammonium bicarbonate (pH 8.5). Myoglob in (3 μg) was added to each well as the internal standard to standardize sample loss during preparation, since commercial isotopic labeled NAPQI modified peptides are not available. Protein denaturation and reduction were done by adding 100 μL of an 8 M urea + 20 mM DTT mixture. Free thiol groups on cysteine residues were then alkylated by adding 100 μL of 20 mM iodoacetamide (IAA) in the dark. The samples were washed twice with 100 μL of ammonium bicarbonate or 100 mM Tris buffer, and then, a new 96-well plate was placed under the filter plate. Samples were digested by MS grade trypsin, chymotrypsin, or Glu-C at a protease/protein (w/w) ratio of 1:25 for 16 h. Digestion was stopped by adding 4 μL of 10% formic acid. Digested peptides were filtered through the filter plate for LC-MS/MS analysis (Scheme 2).

Scheme 2.

High-Throughput Sample Workup Steps for Damaged Protein Analysis

HPLC-MS/MS.

A Thermo Scientific Dionex Ultimate 3000 UHPLC and a Thermo Scientific TSQ Quantiva mass spectrometer system (Thermo Fisher Scientific) were used. For analysis of metabolites, the UPLC column was a C18 reversed phase (Hypersil GOLD, 0.3 × 150 mm, 3 μm, Thermo Scientific) with gradient elution (Table S1). For peptide separations, a Jupiter C18 column (0.5 × 150 mm, 5 μm, Phenomenex) in gradient elution (Table S1) was used. In both LC methods, 98.9% water, 1% ACN (0.1% formic acid) was used as solvent A, and 98.9% ACN, 1% water (0.1% formic acid) was used as solvent B in a gradient elution. The mass spectrometer with XCalibur software (Thermo Scientific) was operated in the positive mode. Product ion scan mode was done using collision-induced dissociation for metabolites and reacted peptide identification. Multiple-reaction monitoring (MRM) was used for metabolite and peptide quantitation. An open-source bioinformatics tool, Skyline,³² was used to generate the mass transition list for MRM (Table S2). MRM data were then imported back to Skyline to obtain peak areas of metabolite and peptides. Detailed HPLC and MS parameters are in the SI file.

Relative Amounts of Reacted Peptides Calculation.

Relative amounts of reacted peptides were estimated as the ratio of peak area of the reacted peptide to that of the internal standard divided by the ratio of the peak area of the unreacted peptide to that of the internal standard in control samples with the assumption that the ionization efficiencies of the unreacted peptide and reacted peptide are similar.

RESULTS

Metabolite Formation.

NAPQI was identified in the absence of hGSTP by trapping with GSH in solution. The product ion spectrum of product NAPQI-GSH at m/z 457.2 shows typical fragmented ion patterns at m/z 328.3, 382.2, and 182.1 corresponding to protonated fragment ions in (Figure 1A).

Figure 1.

LC-MS/MS results for NAPQI-GSH reaction product. (A) Product ion spectrum of NAPQI-GSH conjugate at m/z 457.2 with collision energy of 20 V; (B) extracted ion chromatogram (XIC) for MRM mass transition m/z 457.2 to 328.3 shows peak for NAPQI-GSH at 13.7 min.

Extracted ion chromatogram (XIC) for MRM mass transition m/z 457.2 to 328.3 shows the NAPQI-GSH conjugate eluted at 13.7 min (Figure 1B). These results confirm NAPQI formation in our system. About 10% of acetaminophen converts to NAPQI, but major metabolites are nontoxic glucuronide (acetaminophen-gluc) and sulfate (acetaminophen-sulf) as found in humans.^28,33

Figure 2 shows the relative amounts of NAPQI-GSH found by LC-MRM (details in Figure S1, Table S2) using different enzyme sources. Results were normalized by amounts of enzymes on the magnetic beads estimated by micro bicinchoninic acid (BCA) protein assays,³⁴ Figure S2, concentration of acetaminophen, and time of reaction. The order of bioactivation activity for microsomes was HLM ≈ HLuM > HIM > HKM. The order of bioactivation activity of supersomes was CYPs 1A2 > 2E1 > 3A5 > 3A4 > 2A6 > 1B1.

Figure 2.

NAPQI-GSH conjugate formation with different enzymes. Relative amounts of NAPQI-GSH conjugate {μg of protein}⁻¹ {mM acetaminophen}⁻¹ h⁻¹ found as ratios of area under NAPQI-GSH peak to that of internal standard after incubation for 1 h at 37 °C.

NAPQI-Protein Adducts.

The proteins hGSTP, HSA, and BSA were reacted with metabolites of acetaminophen generated by using magnetic beads coated with cyt P450 enzyme sources. These included human microsomes from liver (HLMs), lung (HLuMs), kidney (HKMs), and intestine (HIMs), cyt P450 (CYP) supersomes CYP1A2 (liver), CYP2A6 (lung), CYP1B1 (kidney), and CYP3A5 (liver and intestine),^35–37 and CYP2E1 and CYP3A4 supersomes active in acetaminophen metabolism.^20,38,39 Experimental workup by Scheme 2 digests enzyme products and provides peptide samples for analysis by HPLC-MS/MS sequencing.^40–42

Figure 3A shows extracted ion chromatograms (XIC) for reacted and unreacted peptides 14–18 and 45–54 and internal standard LFTGHPETLEK. A peptide will increase by the monoisotopic mass of NAPQI, 149.048. The m/z values of unreacted and reacted peptides are listed in Table S3. Reacted peptides CAALR representing amino acids 14 to 18 and ASCLYGQLPK for amino acids 45 to 54 were found in trypsin digested hGSTP. The MS/MS of reacted peptide CAALR with m/z 341.7 (Figure 3B) shows unchanged y₁, y₂, y₃, and y₄ ions and m/z increases of 149 for b₂ and b₃ ions, indicating that Cys-14 reacted. The MS/MS spectrum of reacted peptide ASCLYGQLPK with m/z 614.8 (Figure 3C) shows m/z increases for b₃, b₅, and b₆ ions and increased y₈, y₉ ions indicating reacted Cys-47. The unreacted peptide with a mass increase of 57.021 due to alkylation by iodoacetamide before digestion (peptide 14–18 with m/z 295.7 and 45–54 with m/z 568.8, Table S3) was found in unreacted control samples (Figure S3).

Figure 3.

(A) Extracted ion chromatograms (XIC) for NAPQI-reacted and unreacted CAALR, NAPQI-reacted and unreacted ASCLYGQLPK from hGSTp, and peptide LFTGHPETLEK from internal standard by trypsin digestion; product ion spectrum of (B) NAPQI-reacted peptide CAALR with m/z 341.7; (C) NAPQI-reacted peptide ASCLYGQLPK with m/z 614.8. The b or y ions reflecting the NAPQI modification are indicated in red.

Two MRM transitions were used to identify each peptide, and the one with larger peak area was used for relative quantitation (Figure S4, Table S2). Transitions of m/z 341.7 to 359.2 and 295.7 to 359.2 were used for reacted and unreacted CAALR, m/z transitions 614.8 to 705.4 and 568.8 to 705.4 were used for reacted and unreacted ASCLYGQLPK, m/z 636.3 to 716.4 was used for internal standard LFTGHPETLEK. Relative amounts of reacted peptides were estimated as the ratio of the peak area of the reacted peptide to that of the internal standard divided by the ratio of the peak area of the unreacted peptide to that of the internal standard. Figure 4A shows relative amounts of NAPQI-reacted peptide 14–18 containing Cys-14 with different enzyme sources. Results were normalized by amounts of protein on the magnetic beads (from Micro BCA protein assay, Figure S2), concentration of acetaminophen used, and time of reaction. The order of bioactivation activity of different organ microsomes was HLM > HLuM ≈ HKM > HIM. The order of bioactivation activity of different supersomes was 1A2 > 2E1 > 3A4 > 2A6 > 3A5 > 1B1. For relative amounts of reacted peptide 45–54, the order of the bioactivation activity of different enzymes was similar to that of reacted peptide 14–18 but with 3A5 larger than 3A4 and 2A6 (Figure 4B). The reactivity of Cys-14 was larger than Cys-47.

Figure 4.

NAPQI-reacted peptide formation with different enzymes. Relative amounts of NAPQI-reacted peptide found by LC-MS/MS. (A) CAALR and (B) ASCLYGQLPK {μg of protein}⁻¹, {mM acetaminophen}⁻¹, h⁻¹ formed as a ratio of area under NAPQI-reacted peptide peak to that of internal standard over ratio of area under unreacted peptide peak to that of internal standard in control. CAALR gave more reaction product than ASCLYGQLPK.

Since only two of the thiols on hGSTP were found in the above samples, we explored other digestion enzymes to see if we could detect NAPQI adductions on the other two thiols. Chymotrypsin digestion revealed NAPQI modification on Cys-101. Figure 5A shows extracted ion chromatograms (XIC) for reacted and unreacted peptides 100–103 and internal standard KKHGTVVL. The MS/MS spectrum of reacted peptide 100–103 (RCKY) with m/z 359.7 after chymotrypsin digestion (Figure 5B) shows unchanged y₁, y₂, and b₁ ions and increased b₂, b₃, and y₃ ions, which indicates Cys-101 was reacted. The unreacted peptide 100–103 with m/z 313.7 was found in unreacted control samples. MS/MS spectra of unreacted peptide 100–103 are shown in Figure S5A. Relative amounts of reacted peptide 100–103 were again estimated from peak areas (Table S2 and Figures S5 and S6). The order of bioactivation activity of different organ microsomes was HLM ≈ HLuM > HKM > HIM. The order of bioactivation activity of supersomes was 1A2 > 3A4 > 2E1 ≈ 2A6 > 3A5 > 1B1 (Figure 6A).

Figure 5.

(A) Extracted ion chromatograms (XIC) for NAPQI-reacted and unreacted RCKY and peptide KKHGTVVL from internal standard; (B) product ion spectrum of NAPQI-reacted peptide RCKY with m/z 359.7 by chymotrypsin digestion. The b or y ions reflecting the NAPQI modification are indicated in red.

Figure 6.

NAPQI-adducted peptides formation with different enzymes. Relative amounts of NAPQI-adducted peptide (A) RCKY and (B) VLAPGCLD {μg of protein}⁻¹, {mM acetaminophen}⁻¹, h⁻¹ as ratio of area of NAPQI-adducted peptide peak to that of internal standard over ratio of area under unadducted peptide peak to that of internal standard in control. Relative amount of reacted RCKY is greater than the relative amount of reacted VLAPGCLD.

Glu-C protease digestion reveals a NAPQI modification on Cys-169. Figure 7A shows extracted ion chromatograms (XIC) for reacted and unreacted peptides 164–171 and internal standard IAAKYKE. The MS/MS spectrum of reacted peptide 164–171 (VLAPGCLD) with m/z 468.7 (Figure 7B) shows unchanged y₁, y₂, b₂, b₃, and b₅ ions and increased y₃, y₅, y₆, and b₆ ions, which indicates that Cys-169 reacted. Intact peptide 164–171 with m/z 422.7 was found in unreacted controls (Figure S5B). Relative amounts of reacted peptides 164–171 were estimated from areas of MRM peaks (Table S2; MRM chromatograms in Figure S7). The order of bioactivation activity by organ microsomes was HLuM > HLM > HKM ≈ HIM. The order of bioactivation by supersomes was 1A2 > 2E1 > 2A6 > 3A4 > 3A5 > 1B1 (Figure 6B).

Figure 7.

(A) Extracted ion chromatograms (XIC) for NAPQI-reacted and unreacted VLAPGCLD and peptide IAAKYKE from internal standard, inset plot is the XIC for NAPQI-reacted VLAPGCLD; (B) product ion spectrum of NAPQI-reacted peptide VLAPGCLD with m/z 468.8 by Glu-C digestion. The b and y ions reflecting the NAPQI modification are indicated in red.

The protein HSA and BSA were reacted with NAPQI generated by using magnetic beads coated with cyt P450 1A2. The experimental workup shown in Scheme 2 digests reacted proteins with trypsin and provides peptide samples for analysis by HPLC-MS/MS analysis. Figure 8A shows extracted ion chromatograms (XIC) for reacted peptides from HSA. Figure 8B–D shows MS/MS spectra of reacted peptides LCTVATLR, NECFLQHK, and AACLLPK, indicating the cysteines in the peptides were reacted with NAPQI. Figure S8A–D shows MS/MS spectra of reacted peptides CASLQK, YICENQDSISSK, DVCK, and CCTESLVNR. Figure S9A is the XIC for reacted peptides from BSA. Figure S9B,C shows MS/MS spectra of reacted peptides CASIQK and GACLLPK, indicating the cysteines in the peptides were reacted with NAPQI. Table S4 gives MRM transitions that were used to monitor each reacted peptide.

Figure 8.

(A) XIC for NAPQI-reacted peptides from HSA. Product ion spectra of NAPQI-reacted peptides: (B) reacted LCTVATLR with m/z 513.3; (C) reacted NECFLQHK with m/z 584.3; (D) reacted AACLLPK with m/z 432.6.

The drug raloxifene was also used as a test compound to evaluate our developed method by using magnetic beads coated with cyt P450 3A4. Raloxifene was metabolized to diquinone methide by cyt P450 enzyme and then reacted with cysteines on hGSTP. Figure S10A shows raloxifene-reacted peptides containing Cys-14 and Cys-47. Figure S10B and C show the product ion spectra of reacted peptides indicating the cysteines were modified by raloxifene.

DISCUSSION

The results above demonstrate a new, cell-free in vitro assay protocol that metabolizes test compounds to produce reactive metabolites that react with proteins and enable sample preparation for analysis by LC-MS/MS. This procedure represents a straightforward new test for drug metabolite protein damage that can accommodate a wide range of metabolic and multiple digestion enzymes. The magnetic beads coated with metabolic enzymes act as bioreactors that greatly simplify separation of reacted proteins from the enzyme using a magnet. The 96-well filter plate scheme shortens the sample preparation time to 2 h before digestion and avoids use of a centrifuge. The method is simple, fast, high-enzyme throughput, and facilitates straightforward mapping of reacted positions on proteins from tissue-specific enzymes.

All enzyme sources used here formed the same metabolite for acetaminophen but at different rates. The largest amount of NAPQI formed when human liver microsomes and human lung microsomes were used (Figure 2), which correlates with the liver being identified as the major damage site in acetaminophen overdose.⁴³ Human kidney and intestine microsomes had lower bioactivation rates (Figure 2). Among the supersomes, cyt P450 1A2 produced the largest amount of NAPQI, followed by cyt P450s 2E1 and 3A5. Cyt P450s 1A2, 2E1, and 3A5 are strongly expressed in the human liver,^37,44 and our results are fully consistent with reports that cyt P450s 1A2 and 2E1 are important for acetaminophen bioactivation in humans.^26,45 Cyt P450 3A4 provided moderate bioactivation, while activities of cyt P450s 2A6 and 1B1 were quite low.

Results in Figure 3 show that NAPQI-reacted Cys-14 and Cys-47 on hGSTP were found after trypsin digestion. Cys-47 is highly reactive toward electrophiles due to its relatively low pKa and accessibility.^46,47 Adduction by NAPQI at Cys-14 and Cys-47 was previously reported in vitro,^47–51 and adduction at Cys-47 has been found in cell cultures.⁴⁸ Reacted Cys-101 and Cys-169 were not found using trypsin digestion, since Cys-101 is expected to be recovered in a dipeptide CK, which would have low retention on reverse phase chromatography. Similarly with trypsin, the peptide containing Cys-169 consists of 42 amino acids from residues 141 to 182, too large to be detected.⁴⁹ Chymotrypsin and Glu-C were used to detect reactions on Cys-101 (Figure 5) and Cys-169 (Figure 7). NAPQI on Cys-101 was found in a previous in vitro study using protease Asp-N,⁴⁸ and NAPQI on Cys-169 was reported in an in vitro study using protease pepsin.⁵¹

Our results show that all four cysteines in hGSTP can react with NAPQI, in the ratios of 0.014, 0.00035, 0.017, and 0.0008 for Cys-14 and 47, 101, and 169 using CYP1A2. CYP1A2 gave the highest reaction yields, followed by CYP2E1 (Figures 4 and 6). Among microsomes, human liver microsomes were more efficient at producing NAPQI than other enzymes. Results are consistent with the amounts of NAPQI formation (Figure 2) by the various enzyme sources, i.e., the more metabolite formed, the more reacted peptide. Also, results are consistent with CYP1A2 and CYP2E1 as major enzymes responsible for acetaminophen liver failure.^26,43,45

In our previous work using a similar methodology to form metabolites from test chemicals using CYPs and microsomes, we found that rates of DNA adduction and oxidation correlated very well with Comet DNA damage assays and rodent TD₅₀ genotoxicity assays, as did the reactivities of CYPs and microsomes.^52–54 In the present work, we are only substituting proteins for DNA, so we expect the reactivities of CYPs and microsomes to also correlate with natural systems. We previously showed that CYPs and supersomes provide the same or better activity and the same selectivity when on magnetic beads compared to dispersed in solutions.^55–57

We monitored peptides with one missed cleavage and did not observe signals for miscleaved peptides. If there were miscleaved adducted peptides, the actual adduct level would be higher than that measured. Since we estimated relative amounts of adducted peptides compared with unadducted ones, the ratio ranking of different cytochrome P450 enzymes’ activity would be similar. Also, ionization efficiencies and digestion efficiencies may differ between unadducted and adducted peptides, but the ratio ranking of different enzymes would be the same in all experiments, and results should still provide a nonbiased assessment of relative protein damage from the action of different cytochrome P450 enzymes producing metabolites.

Results of NAPQI reacting with HSA (Figure 8) and BSA (Figure S9) show that eight reacted cysteines in HSA (Cys-98, Cys-124, Cys-200, Cys-223, Cys-288, Cys-339, Cys-499, Cys-500) and four reacted cysteines in BSA (Cys-199, Cys-222, Cys-287, Cys-338) were found by trypsin digestion. Cys-34 in HSA and BSA was reported as the NAPQI binding position in in vivo and in vitro samples by using pepsin digestion.^27,58–60 We did not find reacted Cys-34 in our experiment, since the peptide digested by trypsin containing Cys-34 (ALVLIAFA-QYLQQCPFEDHVK in HSA or GLVLIAFSQYLQQCPF-DEHVK in BSA) is long to have enough signals.

Raloxifene was used as a xenobiotic to further evaluate our developed method for protein modification. Our results (Figure S10) show cysteines Cys-14 and Cys-47 in hGSTP were reacted with raloxifene. Previous works have reported that Cys-47 reacts with raloxifene in vitro,^50,61 and reacted Cys-14 was reported here for the first time.

CONCLUSION

We describe above a new high-throughput sample preparation method coupled with enzyme-coated magnetic bead bioreactors that facilitates bioactivation of drugs to reactive metabolites and reaction with proteins to characterize specific damage to the proteins. Acetaminophen and raloxifene were used as test compounds, and hGSTP, HSA, and BSA were used as target proteins to evaluate the method. All four cysteines in hGSTP (Cys-14, Cys-47, Cys-101, and Cys-169), eight cysteines (Cys-98, Cys-124, Cys-200, Cys-223, Cys-288, Cys-339, Cys-499, Cys-500) in HSA, and four cysteines (Cys-199, Cys-222, Cys-287, Cys-338) in BSA reacted with the acetaminophen metabolite. Two cysteines (Cys-14, Cys-47) in hGSTP were found to react with raloxifene metabolite. This new methodology is suitable for organ-specific analysis of protein damage by metabolites of drugs and other chemicals. The method provides exact sites and relative amounts of adduct formation compared with traditional radioisotope or immunological techniques and also identifies the most active enzymes for active metabolite generation.