Identification of Human Liver Microsomal Proteins Adducted by a

Article pubs.acs.org/crt

Identification of Human Liver Microsomal Proteins Adducted by a Reactive Metabolite Using Shotgun Proteomics Yanou Yang,* Qing Xiao, W. Griffith Humphreys, Ashok Dongre, and Yue-Zhong Shu Bristol-Myers Squibb Research and Development, 311 Pennington-Rocky Hill Road, Pennington, New Jersey 08534, United States S Supporting Information *

ABSTRACT: Covalent modification of cellular proteins by chemically reactive compounds/metabolites has the potential to disrupt biological function and elicit serious adverse drug reactions. Information on the nature and binding patterns of protein targets are critical toward understanding the mechanism of drug induced toxicity. Protein covalent binding studies established in liver microsomes can quantitively estimate the extent of protein modification, but they provide little information on the nature of the modified proteins. In this article, we describe a label-free shotgun proteomic workflow for the identification of target proteins modified in situ by reactive metabolites in human liver microsome incubations. First, we developed a shotgun proteomic workflow for the characterization of the human liver microsomal subproteome, which consists of predominately membrane-bound proteins. Human liver microsomes were solubilized with a combination of MS-compatible organic solvents followed by protein reduction, alkylation, and tryptic digestion. The unmodified samples were analyzed by UHPLC-MS/MS, and the proteins were identified by database searching. This workflow led to the successful identification of 329 human liver microsomal subproteome proteins with 1% FDR (false discovery rate). The same method was then applied to identify the modifications of human liver microsomal proteins by a known reactive metabolite 2-(methylsulfonyl)benzo[d]thiazole (2), either after incubation directly with 2 or with its parent compound 2-(methylthio)benzo[d]thiazole (1). A total of 19 modified constituent peptides which could be mapped to 18 proteins were identified in human liver microsomes incubated directly with 2. Among these, 5 modified constituent peptides which could be mapped to 4 proteins were identified in incubation with 1, which is known to generate 2 in human liver microsomal incubations. This label-free workflow is generally applicable to the identification and characterization of proteins adducted with reactive metabolites in complex matrices and may serve as a valuable tool to understand the link between protein targets and clinically relevant toxicities.

■

INTRODUCTION Covalent modification of cellular proteins by chemically reactive drug metabolites has the potential to alter critical cellular pathways and trigger autoimmune responses, which can lead to clinical adverse events such as drug induced liver injury (DILI) or drug hypersensitivity.1,2 Protein covalent binding studies in liver microsomes with radiolabeled compounds have been an integral part of drug candidate evaluation in the pharmaceutical industry and may affect the progression of a drug candidate.3,4 However, recent studies have indicated that the amount of protein covalent binding alone is not predictive of toxicity.3−7 One hypothesis is that reactive metabolites might have different reactivity to critical target proteins and that their different binding patterns lead to distinct biological responses.8 In order to better understand the mechanism of drug induced toxicity caused by reactive metabolites, it is of great interest to comprehensively characterize reactive metabolite−protein adducts at the amino acid level so that critical protein damage can be distinguished from noncritical ones.2,9 Despite the great interest, detection and identification of protein adducts in complex mixtures are challenging tasks since protein adducts are typically present at low abundance with a © 2014 American Chemical Society

very wide dynamic range and a large excess of nonadducted proteins.10,11 Although protein covalent binding studies in liver microsomes using radiolabeled compounds can quantitively estimate the extent of protein modification, it provides little information on the nature of the modified proteins.4,12 The traditional radiochemical and immunochemical approaches can trace down the covalent modification at protein level, but little is known about which amino acids on the adducted proteins get modified.13,14 Therefore, there is great need to develop sensitive and selective analytical methodologies for the detection and identification of drug−protein adducts at amino acid level in complex biological samples. The introduction of shotgun proteomics has enabled the rapid analysis of complex protein samples.15,16 In this bottomup approach, the identities of proteins in a complex mixture are determined by digestion of the proteins into peptides prior to analysis. Rigorous LC separation of peptides has critical importance in reducing ion-suppression effects and increasing the peptide coverage and thus the number of identified Received: May 11, 2014 Published: August 8, 2014 1537

dx.doi.org/10.1021/tx500181p | Chem. Res. Toxicol. 2014, 27, 1537−1546

Chemical Research in Toxicology

Article

sequencing-grade modified trypsin was purchased from Promega (Madison, WI), and formic acid, HPLC grade water, and acetonitrile were purchased from EMD chemicals (Gibbstown, NJ, USA). Solubilization and Digestion of Human Liver Microsomes. Human liver microsomes (50 μg, 2.5 μL of 20 mg/mL stock) in duplicate were added to 25 μL of the following different solubilizing solutions: (1) 50% TFE in 50 mM ammonium bicarbonate buffer containing 5 mM DTT (labeled as TFE); (2) 60% MeOH in 50 mM ammonium bicarbonate buffer containing 5 mM DTT (labeled as MeOH); (3) 50 mM ammonium bicarbonate buffer containing 5 mM DTT (labeled as Direct). All samples were solubilized at 50 °C for 60 min. The proteins were then alkylated with IAM at a final concentration of 10 mM at room temperature in the dark for 45 min. Samples were diluted 10 times with 50 mM ammonium bicarbonate buffer followed by digestion with trypsin. Trypsin digestion was carried out using sequencing-grade modified trypsin 1:25 trypsin-to-protein ratio w/w at 37 °C overnight. Digested samples were stored at −80 °C until LC-MS/MS analysis. Incubation of Compounds with Human Liver Microsomes. Compound 1 or 2 at 30 μM was incubated in duplicate for 1 h at 37 °C with human liver microsomes in 100 mM phosphate buffer (pH 7.4) in the presence of NADPH (2 mM). Control incubations in the absence of compound were also done in duplicate. The protein concentration of liver microsomes used in the incubation was 1 mg/ mL. The total incubation volume was 1 mL. Incubation samples (100 μL) were taken at 0 and 60 min and quenched with an equal volume of acetonitrile followed by centrifugation. The supernatant was transferred for UHPLC-UV/MS analysis to conduct metabolite profiling. Another two aliquots (50 μL) of each incubation sample were ultracentrifuged at 100,000g for 1 h. The microsome pellets were resuspended in 25 μL of TFE or MeOH solubilization solutions. The resuspended samples were solubilized, alkylated, and digested by following the same procedures as those described in the Solubilization and Digestion of Human Liver Microsomes section above. UHPLC-UV/HRMS Methods for Metabolite Profiling. The metabolite profiling of the supernatant from the incubation samples were analyzed using an Accela UHPLC system coupled to a LTQOrbitrap mass spectrometer (Thermo Fisher Scientific, San Jose, CA). Chromatographic separations were performed with an Acquity UPLC High Strength Silica T3 1.8 μm (2.1 mm × 100 mm) column at a flow rate of 0.6 mL/min. The column was eluted with water containing 0.1% formic acid (solvent A) and acetonitrile containing 0.1% formic acid (solvent B). The UHPLC separation used a linear gradient program of 5% B for 0.5 min, 5−70% B for 8.5 min, 70−95% for 1.5 min and 5% B for 1.5 min. The column and samples were maintained at temperatures of 55 and 5 °C, respectively. The UV data were collected for a scan range of 200−600 nm using an in-line Accela PDA detector (Thermo Fisher Scientific, San Jose, CA). The HPLC eluent was directly coupled to an LTQ/Orbitrap mass spectrometer (Thermo Fisher Scientific). Mass spectrometric analysis was conducted with an electrospray ionization source. Mass calibration was performed daily according to the manufacturer’s guidelines. The capillary temperature was 275 °C, the capillary voltage was 34 V, the source voltage was 4 kV, and the tube lens voltage was 108 V. Three scan events were used as follows: (1) m/z 100−1000 full-scan MS operated at a target mass resolution of 15000 (full-width at half-maximum as defined at m/z 400), (2) data-dependent MS2 scan with a mass resolution of 7500 on the most intense ion from the full-scan event, and (3) data-dependent MS3 scan with a mass resolution of 7500 on the most intense ion from the MS2 scan. All data were processed in the Qual browser module of Xcalibur (ThermoFisher Scientific). UHPLC-UV/HRMS Methods for Peptide Analysis. All digested samples were analyzed using an Accela UHPLC system coupled to an LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). The injection volume was 50 μL, which corresponds to 10 μg of starting proteins. Chromatographic separations were performed with an Acquity UPLC BEH 130 C18 1.7 μm (2.1 × 100 mm) column at a flow rate of 0.6 mL/min. The column was eluted with water containing 0.1% formic acid (solvent A) and acetonitrile containing 0.1% formic acid (solvent B). The UHPLC separation used a linear

proteins. While reversed phase nanoflow high performance liquid chromatography (nHPLC) has been the method of choice for peptide separation in many global proteomics analysis, ultrahigh-performance liquid chromatography (UHPLC) is increasingly used in shotgun proteomics analysis.17,18 A recent study indicated that the standard flow UHPLC-based platform can be used with no loss of sensitivity and that this platform is preferable in peptide analysis because of robustness, ease of use, and multiplexing capability, provided that there is sufficient sample for analysis.19 Recently, bottom-up protein analysis has been successfully used to trace down the covalent binding of reactive metabolites to certain amino acids for the proteins of interest, including model target proteins with known reactivity,20−23 cytochrome P450s (P450),24−27 and microsomal proteins in rat liver microsomes.28 Nevertheless, the analysis in these studies either has been limited to characterizing the covalent modification of a single protein or requires a radiolabeled compound. Although human liver microsomes are widely used to study in vitro human drug metabolism and are commonly used in protein covalent binding studies to generate human relevant reactive metabolites,4,12 no study on the LC−MS characterization of covalent modification of proteins by reactive metabolites generated in situ in human liver microsomes has been reported thus far. One of the main reasons could be that human liver microsomes consist of complex mixtures of membrane-bound proteins, which would make the detection of protein adducts using shotgun proteomics very challenging.29,30 The objectives of present study include (1) to establish sample preparation protocols for bottom-up analysis of microsomal proteins in human liver microsomes; and (2) to apply the established proteomic workflow to detect/identify reactive metabolite−protein adducts at amino acid level in human liver microsomes. To ensure peptide coverage, we utilized a combination of different MS-compatible membrane solubilization protocols in sample preparation and developed a UHPLC-Orbitrap method for optimal peptide separation/ detection. The model compound used in this study is 2(methylthio)benzo[d]thiazole (1 in Figure 1), which has been

Figure 1. Structures of model compounds.

shown to form the highly reactive metabolite 2(methylsulfonyl)benzo[d]thiazole (2 in Figure 1) and a significant amount of GSH adduct in human liver microsomal incubations.31 The generic label-free shotgun proteomic workflow described in the current article can be used to characterize drug−protein adducts in other complex biological matrices and ultimately help us better understand drug induced toxicity caused by reactive metabolites.

■

EXPERIMENTAL SECTION

Materials. human liver microsomes were obtained from BD Gentest (150 donor mixed gender pool, Woburn, MA). NADPH, methanol (MeOH), trifluoroethanol (TFE), ammonium bicarbonate, dithiothreitol (DTT), and iodoacetamide (IAM) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Compounds 2-(methylthio)benzo[d] thiazole (1) and 2-(methylsulfonyl)benzo[d]thiazole (2) were purchased from Ryan Scientific, Inc. (Mt. Pleasant, SC), 1538



Article

Table 1. Comparison of Different Solubilization Conditions for the Identification of Peptides/Proteins in Human Liver Microsomes Using Proteomics Approaches samplea

# of proteins in replicate 1

# of proteins in replicate 2

Reproducibility (%)b

# of proteinsc

# of peptides

average % sequence coveraged

TFE MeOH Direct

306 307 110

310 306 160

94 92 55

317 319 181

2097 2236 420

22.6 24.4 7.7

a

Sample treatment conditions are explained in the Experimental Section section. bRatio between number of proteins identified in both technical replicates and total number of proteins identified under the corresponding treatment condition. cData are reported based on proteins identified under each treatment condition with 1% FDR. dData are reported based on the averaged sequence coverage from all the proteins identified under each specified condition. gradient program of 2% B for 1 min, 2−35% B for 49 min, 35−80% for 7 min, and 2% B for 3 min. The column and samples were maintained at temperatures of 55 and 5 °C, respectively. The UHPLC eluent was directly coupled to an LTQ/Orbitrap mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). The LTQ Orbitrap MS was operated in the positive ion electrospray mode using nitrogen as the sheath and auxiliary gases. The heated capillary temperature was set at 330 ◦C, the capillary voltage was 20 V, the source voltage was 5 kV, and the tube lens voltage was 140 V. Each sample was injected twice. For both injections, the acquisition cycle consisted of a full scan in the orbitrap with a mass range from m/z 350 to 2000 at the resolution of 60,000, followed by 6 data-dependent MS/MS scans acquired in the linear ion trap using collision-induced dissociation (CID). For the first injection, the six most intense ions at a threshold above 500 was selected for CID, while the top 7−12 most intense ions at a threshold above 200 was selected for CID in the second injection. Precursor ions that were detected twice within 10 s were put on a dynamic exclusion list for a period of 30 s. The CID-MS/MS collision energy was set at 35%. The charge state screening was enabled with +1 and unassigned charge states rejected for MS/MS scans. Xcalibur 2.0 was used for both data acquisition and data processing. Extracted ion chromatograms (XICs) were generated with a 10 ppm mass window centered on the exact m/z of peptides of interest. Quan Browser within Xcalibur was used to perform the peptide MS peak integration and peak area reporting. Database Searching. The acquired MS/MS spectra were searched against the refseq human database using Sorcerer-SEQUEST software (Sage-N Research Inc. Milpitas, CA). The adduction of reactive metabolites as well as S-carbamidomethylation of Cys (+57.02146 Da) and oxidation of methionine (+15.99492 Da) were specified as variable modifications. The mass shift produced by adduct formation was 132.9986, which is based on the mass shift of GSHtrapped reactive metabolite formed in human liver microsomal incubation as reported previously.31 Less than two missed cleavages by trypsin were also considered in the search. Peptide mass tolerance was set at ±15 ppm. Fragment mass tolerance was set at ±1.0 Da. Scaffold (version Scaffold_4.0.6.1, Proteome Software Inc., Portland, OR) was used to validate MS/MS based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 95.0% probability by the Peptide Prophet algorithm32 with Scaffold delta-mass correction. Protein identifications were accepted if they could be established at greater than 99.0% probability and contained at least 1 identified peptide. Protein probabilities were assigned by the Protein Prophet algorithm.33 Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. For the identified proteins, the Gene Ontology term for endoplasmic reticulum (ER) was fetched using in-house software.

fully accessible for digestion and achieve high protein sequence coverage for membrane bound microsomal proteins. A critical step in analyzing membrane proteins in shotgun proteomic approaches is dissociation and solubilization of lipid bilayerbound proteins from the membrane. Different solvent groups have been used to promote the solubilization of membrane proteins, but a majority of them either alter proteolytic digestion or complicate LC-MS/MS analysis, requiring an additional sample cleanup step postdigestion. Two organic solvents,34 methanol (MeOH) and 2,2,2-trifluoroethanol (TFE), have been recently recommended for miscible extraction of membrane proteins in shotgun proteomics since sample cleanup could be avoided leading to lower sample loss and better recovery of low abundance, highly hydrophobic transmembrane peptides. In order to select the solubilization approach with highest microsomal proteome coverage, we treated human liver microsomes with three different conditions: (1) solubilization with 50% TFE, (2) solubilization with 60% MeOH, and (3) direct digestion without prior solubilization step. Table 1 gives the compiled results of all three conditions from this study in terms of total number of peptides/proteins identified, reproducibility, and average sequence coverage. As expected, including a solubilization step with either MeOH or TFE led to much higher numbers of proteins/peptides identified with superior reproducibility and better sequence coverage. The Venn diagrams in Figure 2 compare the numbers of peptides/ proteins identified under each treatment condition. Combining all three conditions, a total of 2624 peptides were identified at greater than 95.0% probability in human liver microsomes (Supporting Information, S1), with 408 and 291 peptides

■

RESULTS AND DISCUSSION Identification of Proteins in Human Liver Microsomes Using Different Solubilization Approaches. In order to identify covalent binding sites of reactive metabolites on proteins in a complex protein mixture like human liver microsomes, it is very important to render the protein mixture

Figure 2. Venn diagram representing the number of identified proteins (A) or peptides (B) by different solubilization approaches. 1539



Article

most of the proteins listed in Table 2. The rank order of percentage coverage for the individual enzymes is consistent with their abundance in the human liver.35 This suggests that the solubilization with MeOH and TFE is nondiscriminative. Among the 329 proteins identified, 114 proteins were known to be located in the ER. According to the preparation procedures for microsomal subcellular fraction preparation,36,37 it is not surprising that proteins from other organelles, such as mitochondria and cytoskeleton, are also present in the human liver microsomes. Since we intended to leverage UHPLC-MS instrumentation already available in a typical drug metabolism laboratory for initial investigations, we did not use nanoflow LC, which is the mainstream method for global proteomics analysis. The number of microsomal peptides/proteins (2624/329) detected using our analytical platform is in range with those reported in the literature using nanoflow LC.34,38 This is consistent with previous findings that the standard flow UHPLC-based platform can be used with no loss of sensitivity in comparison to nanoflow LC.19 Identification of Modified Peptides and Corresponding Microsomal Proteins. After establishing a workflow for identifying proteins in human liver microsomes through shotgun proteomics combining two MS-compatible solubilization approaches as described above, the next goal was to apply this workflow to detect/identify modified peptides and thus the corresponding proteins in human liver microsomes incubated with a compound, which is known to generate reactive metabolites. Compound 1 was previously found to be metabolized in human liver microsomal incubation in the presence of NADPH to highly reactive sulfone metabolite, 2, which can form significant amounts of GSH adduct by nucleophilic displacement in the presence of GSH.31 If GSH is not present in the incubation, 2 formed in the human liver microsomal incubation was anticipated to modify microsomal proteins through cysteine adduction (Figure 3).

identified unique to using MeOH and TFE, respectively. Among the 2624 peptides identified, 416 of them were cysteine containing peptides. The total number of proteins indentified with 1% FDR in human liver microsomes was 329 (Supporting Information, S2), with 11 and 9 proteins unique to using MeOH and TFE, respectively. Although 75 unique peptides were identified under direct digestion conditions, the identification of these unique peptides did not lead to identification of additional unique proteins, and the combination of MeOH and TFE covers 100% of the proteins identified. However, the MeOH and TFE solubilization approaches are complementary to each other. Therefore, a combination of both MeOH and TFE approaches are later used to solubilize human liver microsomes incubated with compounds for the identification of modified peptides, while direct digestion was not further evaluated. Three important groups of drug metabolism enzymes, including cytochrome P450s, cytochrome P450 reductase and cytochrome b5, microsomal glutathione S-transferases (GSTs), and UDP-glycosyltransferase (UGTs), are known to be abundant in liver microsomes. We compared the sequence coverage and number of unique peptides identified under three different solubilization conditions for individual proteins from these three groups of enzymes, and the results are shown in Table 2. Consistent with the overall performance shown in Table 1, using MeOH or TFE increased the number of unique peptides and sequence coverage dramatically for all the proteins listed in the table. Although TFE and MeOH have similar performance, MeOH has slightly better sequence coverage for Table 2. Comparison of Different Solubilization Approaches for Sequence Coverage (%) and Number of Unique Peptides (in Parentheses) for Selected Proteins of Interest in Human Liver Microsomes protein name (human)

accession #

TFE

MeOH

P450 3A4

NP_059488.2

P450 2C9

NP_000762.2

P450 2E1

NP_000764.1

P450 reductase

NP_000932.3

P450 2A6

NP_000753.2

cytochrome b5 P450 2C8

NP_683725.1 NP_000761.3

P450 1A2 GST A1 microsomal GST 1 microsomal GST 2 microsomal GST 3 UGT 2B7 UGT 2B4 UGT 1A4

NP_000752.2 NP_665683.1 NP_665707.1 NP_002404.1 NP_004519.1 NP_001065.1 NP_066962.1 NP_009051.1

UGT UGT UGT UGT

NP_066307.1 NP_000454.1 NP_001066.1 NP_001063.2

44.5% (16) 44.5% (13) 40.0% (14) 35.5% (16) 37.5% (14) 63.0% (6) 27.5% (10) 19.5% (6) 47% (9) 38% (5) 31.0% (2) 8.6% (1) 41% (17) 45% (12) 30.5% (12) 28.5% (5) 27.5% (5) 27% (5) 17% (1)

47.5% (16) 47.5% (15) 49.5% (18) 45.5% (21) 43.0% (17) 77.0% (6) 33.5% (13) 22.5% (8) 45.0% (8) 63.0% (6) 20.3% (2) 16.8% (2) 46% (18) 46% (14) 34.5% (14) 33% (7) 31.5% (7) 25% (4) 18.5% (2)

a

1A9 1A1 2B10 1A6

Direct 13% (6) 13.0% (4) 12.1% (5) 11.0% (5) 13.3% (5)

Figure 3. Adduction reaction scheme.

31.5% (3) 10.7% (4)

In order to verify that we can use the established human liver microsome proteomics workflow to identify protein modifications by the reactive metabolite, we first incubated 2 directly with human liver microsomes in duplicate in the absence of NADPH since 2 is intrinsically reactive without cytochrome P450-mediated bioactivation. Human liver microsomal proteins in the incubation were then pelleted by ultracentrifugation and prepared by solubilization, alkylation, and digestion procedures as established with untreated human liver microsomes. Since the incubations were done in duplicate and two solubilization approaches were used in sample preparation, a total of four compound 2 samples were generated for analysis. The tryptic peptide mix of these four samples were analyzed by LC-MS/ MS followed by targeted database searching to specifically look for peptides with the addition of C7H3NS (+132.9986) to

1.6% (1) 3.0% (1) 8.0% (1) -a 6.5% (3) 4.05% (1) 3.75% (1) -

-, no unique peptide was identified. 1540


1541

6 1 27 42 2 5 6 5 4

1405.5431 1508.6428 1528.6479 1656.7428 1540.6612 1659.7214 1662.8473 2034.9867 1322.5722 3306.4316 (3+)

R.GFEGSFEELCR.N R.VTCIDPN PNFEK.F K.VFANPEDCVAFGK.G* R.KVFANPEDCVAFGK.G* R.DLTDCLLVEMEK.E R.SIQFVDWCPTGFK.V K.VACITEQVLTLVNK.R* R.VLVVGLGNSGCDIATELSR.T R.LGMFNIQHCK.K

R.CPEALFQPSFLGMESCGIHETTFNSIMK .C

(2+) (2+) (2+) (2+, 3+) (2+) (2+) (2+) (2+) (3+) 1

43.7

37.0 28.5 33.6 31.0 41.3 41.9 40.3 39.6 29.4

38.3

45.1 28.7 27.9 35.2

28.5 33.9 35.7 35.0

retention time (min)e

MeOH MeOH, MeOH MeOH, MeOH, MeOH MeOH, MeOH, TFE MeOH,

−8.3 −2.2 −1.0 −4.4 −4.6 −4.3 −7.0 −3.7 −0.5 −3.9 MeOH

MeOH MeOH, TFE MeOH MeOH, TFE

−6.7 −2.4 −4.0 −4.0

2.1

MeOH, TFE MeOH MeOH, TFE MeOH, TFE

−5.6 0.8 −3.8 −4.8

TFE

TFE TFE

TFE TFE

TFE

solubilization methodg

mass accuracy (ppm)f

P450, 2E1 (NP_000764.1) tubulin, alpha-4A chain (NP_005991.1) ribophorin I precursor (NP_002941.1) flavin containing monooxygenase 3 (NP_001002294.1) serine (or cysteine) proteinase inhibitor, clade A (alpha-1 antiproteinase, antitrypsin), member 1 (NP_000286.3) beta actin (NP_001092.1)

41,710/79.2

56,812/51.1 49,892/43.8 68,527/61.3 59,994/69.4 46,737/39.2

77,893/57.0 28,301/37.7 17,587/47.0

(266) (98) (246) (98) (98) (221)

(268) (347) (477) (197) (256) 1 (272)

1 1 1 1 1

1 (625) 1 (96) 1 (50)

1 1 1 1 1 1

1 (159)

23,804/25.6 55,789/42.7 57,306/49.3 40,196/60.5 39,842/48.0 39,810/53.9 38,132/59.5

1 (129)

21,658/51.8

progesterone receptor membrane component 1 (NP_006658.1) progesterone receptor membrane component 2 (NP_006311.1) P450 2C8 (NP_000761.3) P450 3A4 (NP_059488.2) alcohol dehydrogenase 4 (NP_000661.2) alcohol dehydrogenase 1C (NP_000660.1) alcohol dehydrogenase 1B (NP_000659.2) D-beta-hydroxybutyrate dehydrogenase, mitochondrial (NP_004042.1) long-chain-fatty-acid-CoA-ligase I (NP_001986.2) methyltransferase like protein 7A (NP_054752.3) microsomal glutathione S-transferase 1 (NP_665707.1)

3 (255, 600, 761)

# of modified Cys (position)

164,835/58.5

MW (Da)/ protein coverage (%)i

carbamoyl-phosphate synthetase 1, mitochondrial (NP_001866.2)

name (accession number)

h

protein

Modified cysteines are underlined. bCalculated monoisotopic elemental mass. The observed monoisotopic elemental mass is within 10 ppm of calculated values. cObserved charge state. dTotal number of spectra detected for the same modified peptide from all four samples. eAverage retention time if multiple spectra were detected for the same modified peptide. fAverage mass accuracy when multiple spectra were detected for the same modified peptide. gSolubilization methods under which the modified peptides were detected. hThe accession number is from the Reference Sequence (RefSeq) collection. iOn the basis of the number of covered amino acid vs total number of amino acids in the sequence. Peptide sequences labeled with * were also detected in incubation with 1 after bioactivation.

a

2

1375.5689 (2+)

2 2 4 2

K.FGVEAFSDCLR.Y

(2+) (2+) (2+) (2+)

1245.5562 1250.4848 1262.5536 1334.6515

5 1 9 2

R.DFIDCFLIK.M* K.ECYSVFTNR.R K.ALGATDCLNPR.D K.VIPLFTPQCGK.C

(2+) (2+) (2+) (2+)

# of detected spectrad

1035.4881 1667.7548 1823.7568 1099.4831

monoisotopic elemental massb (charge)c

K.VVAVDCGIK.N R.SAYALGGLGSGICPNR.E K.TSACFEPSLDYMVTK.L* R.GLATFCLDK.D

sequence

a

peptide

Table 3. Summary of Covalently Modified Peptides/Proteins Detected in Human Liver Microsomes Incubated with 30 μM Compound 2

Chemical Research in Toxicology Article



Article

Figure 4. MS data from incubation with 30 μM 2 for peptides VFANPEDCVAFGK and TSACFEPSLDYMVTK modified by the reactive metabolite. (A) Full scan accurate mass acquired in the orbitrap; (B) CID fragment ion spectra acquired in the ion trap; (C) proposed main CID fragmentation route.

database search is supported by the CID fragmentation pattern (Supporting Information, S5−S23). The data from two modified peptides VFANPEDCVAFGK and TSACFEPSLDYMVTK are shown as examples (Figure 4). In order to ascertain whether the adducted peptides were compound related, we checked the presence of these peptides in the control human liver microsomal incubations where the compound was absent. None of these adducted peptides was detected in the control incubation samples. Instead, the IAM adducted variations of these peptides, which correspond to the unmodified peptides, were all detected in control incubation

cysteine residues. The peptide probability requirement was to be equal to or above the value which gives 1% FDR in peptide identification. On the basis of these criteria, a total of 19 modified peptides were identified in the incubation with 2. The sequences and monoisotopic elemental masses of these peptides are shown in Table 3. For each modified peptide, there were often multiple mass spectral data sets as supporting evidence from a combination of four samples. The mass accuracy for the assigned modified peptides is less than 15 ppm, and most of them are well below 10 ppm (Supporting Information, S3). Each modified peptide identified by the 1542



Article

Figure 5. LC-UV chromatograms of 1 and 2 incubated with human liver microsomes in the presence of NADPH. (A) 1, 0 min; (B) 1, 60 min; (C) 2, 0 min.

samples. Furthermore, since adducting with the reactive metabolite increases the hydrophobility of the peptides, the reactive metabolite adducted peptide variations should elute later than the corresponding IAM modified variations. Indeed, we observed that the reactive metabolite modified variations eluted 7−17 min later than the corresponding IAM modified variations (Supporting Information, S4). The search algorithm for modified peptides used here is the same as identifying posttranslational modifications (PTM). The amino acid to be modified and its molecular weight change from the reactive metabolite adduction have to be defined in the search. This targeted search algorithm requires prior knowledge of the reactive metabolite and its formation mechanism, as for 2 in this study. We are currently exploring a nontargeted search algorithm for identifying modifications to unknown amino acid residues by unknown reactive metabolites. According to sequences of identified peptides, we can then determine which proteins are most likely to be present in the sample (Table 3). These proteins have to meet the criteria of being present at greater than 99.0% probability and contain at least 1 unique peptide. As shown in Table 3, multiple modified peptides identified in the sample can be associated with one single protein. For example, three modified peptides were determined to come from mitochondrial carbamoyl-phosphate synthetase 1 with three different modification sites. While for microsomal glutathione S-transferase 1, two different peptides were identified but correspond to the same cysteine modification. However, one modified peptide can be associated with different but closely related proteins due to protein identification ambiguity, as in the case of peptides R.GLATFCLDK.D and K.VIPLFTPQCGK.C.

After successfully identifying protein/peptide modifications by the reactive metabolite using the established workflow, we incubated 1 with human liver microsomes in the presence of NADPH. As shown in Figure 5, 1 completely converted to two metabolites at retention times 4.1 min (79% of total UV peak area) and 4.8 min (21% of total UV peak area) after 1 h of incubation. MS fragmentation and retention time matching with synthetic standards confirmed that the two metabolites formed from incubation with 1 are the sulfoxide and sulfone (2) metabolites (data not shown). These results are consistent with previous results.31 After confirming the conversion of 1 to reactive sulfoxide and sulfone metabolites, human liver microsomes from the incubation was prepared and analyzed the same way as that previously described for identifying modified peptides. Five modified peptides were identified in the incubation with 1, all of which were previously identified in incubation with 2 (labeled with an asterisk in Table 3). However, 14 peptides which were detected in incubation with 2 were not detected in incubation with 1. This can be attributed to a much lower concentration of 2 formed in the incubation with 1. The concentration of 2 present in the incubation samples at the end of 1 h of incubation was approximately 6 μM (estimated by UV), which is much lower than the 30 μM used in the incubation with 2. Furthermore, 2 was gradually formed during the incubation with 1, and the average concentration of 2 over the course of the incubation is even lower, leading to a lower percentage of modified proteins and a lower concentration of modified peptides after digestion, which may fall below the detection limit of the system. 1543



Article

Figure 6. Comparison of extracted ion chromatograms of peptide VFANPEDCVAFGK modified by the reactive metabolite (m/z = 623.7854, doubly charged) or IAM (m/z = 585.7968, doubly charged) after incubating with 1 at 30 μM (A and B) and 2 at 30 μM (C and D).

Table 4. Relative Ratio of Modified and Unmodified Peptides % adductiona (±SDb) peptide sequence R.DFIDCFLIK.M K.VFANPEDCV AFGK.G R.KVFANPED CVAFGK.G K.VACITEQV LTLVNK.R K.TSACFEPS LDYMVTK.L

adduction

m/z (charge)

retention time (min)

30 μM, 1

30 μM, 2

reactive metabolite adduction carbamidomethylation reactive metabolite adduction carbamidomethylation reactive metabolite adduction carbamidomethylation reactive metabolite adduction carbamidomethylation reactive metabolite adduction carbamidomethylation

623.7854(2+) 585.7968 (2+) 765.3312(2+) 727.3427(2+) 553.2549 (3+) 527.9292(3+) 832.4309 (2+) 794.4424 (2+) 912.8857 (2+) 874.8971 (2+)

45.2 33.7 33.8 21.3 31.2 19.6 40.4 31.2 35.8 26.2

3.2 ± 1.2

16.5 ± 2.2

13.4 ± 1.1

41.7 ± 1.6

8.8 ± 2.7

36.6 ± 3.4

16.2 ± 3.5

42.7 ± 6.3

15.0 ± 2.9

42.6 ± 5.4

a

The percentage was calculated by dividing the MS peak area of the reactive metabolite-modified peptide by the combined MS peak area from the same peptide with both reactive metabolite and IAM modifications. bn = 4.

unmodified portion, and the combination of reactive metabolite-adducted and the IAM adducted variation for the same peptide reflects the total peptide. Assuming MS responses from these two peptide variations are the same, the MS peak area in the extracted ion chromatograms of a given peptide can be used to estimate the adduction percentage. This MS-based approach has been used previously. 39 Using peptide VFANPEDCVAFGK as an example (Figure 6), the calculated percentage of peptide adducted with the reactive metabolite is 14.5% and 39.8% for incubation with 1 and 2 respectively. A total of four samples were analyzed from incubation with each of 1 and 2, while Figure 6 only shows data from one sample

If our assumption on the concentration dependency is true, we expect to see a higher percentage of the same peptide adducted with the reactive metabolite in the incubation with 2 than with 1. In order to verify this, we need to compare the fraction of a given peptide modified by reactive metabolite in incubation with 1 to that in incubation with 2. Since the postincubation human liver microsomes went through solubilization/denaturation and alkylation steps, any cysteines which were not adducted with the reactive metabolite during the incubation with compound would have been adducted with alkylating reagent IAM. Therefore, the portion of a given peptide with carbamidomethylation on cysteine reflected the 1544



Article

in complex matrices. Importantly, our approach can be relatively readily adapted by a typical drug metabolism laboratory and can thus serve as a valuable tool to better understand the mechanism of drug induced toxicity with potential association to reactive metabolite(s).

from each compound incubation. The data from all four samples incubated with 1 or 2 for the five modified peptides identified in incubation with 1 are summarized in Table 4. For all five peptides, the percentage of reactive metabolite adduction in incubation with 2 is 3−5-fold of that in the incubation with 1. One concern with calculating the percentage of modification using MS peak area was that the MS ionization efficiency for reactive metabolite-adducted and IAM adducted variations for the same peptide might be different. In order to validate our data, we normalized the MS peak area of the carbamidomethylated peptides (unmodified portion) with total MS peak area of all peptides detected for the sample, then compared the normalized MS peak area of the carbamidomethylated peptides in human liver microsomes incubated with 1 and 2 to that of control human liver microsomes without incubating with any compound. For all five peptides, the decrease of unmodified peptide is always higher in human liver microsomes incubated with 2 than with 1 (data not shown), which is consistent with data in Table 4. This data clearly indicates the percentage of protein modified by the reactive metabolite is dependent on both the inherent reactivity and the amount of the reactive metabolite present in the incubation, even in the case where there is already excess amount of reactive metabolite present. The 30 μM compound concentration used in our work is in range with those (10−50 μM) used in the literature for similar type of work. However, such a concentration range is not entirely unreasonable when one considers liver exposure and hepatic metabolism of orally absorbed compounds, known as “first pass effect.” We realize that the concentration might be on the high side for some in vivo situations, but we believe the sensitivity of this workflow can be improved by including additional sample preparation steps, such as protein enrichment and fractionation, and using longer chromatographic analysis time. We intend to explore those aspects in our future work.

■

ASSOCIATED CONTENT

S Supporting Information *

List of peptides and proteins identified in human liver microsomes, list of modified peptides identified in human liver microsomes incubated with compounds 1 or 2, retention time change after the peptides were adducted, spectra, and fragmentation tables of modified peptides. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Tel: 01-609-818-6332. Fax: 01-609-818-7877. E-mail: yanou. [email protected]. Funding

This study was fully financed by Bristol Myers Squibb Company. Notes

The authors declare no competing financial interest.

■

ABBREVIATIONS UHPLC/HRMS, ultrahigh pressure liquid chromatography/ high resolution accurate mass spectrometry; FDR, false discovery rate; P450, cytochrome P450; MeOH, methanol; TFE, trifluoroethanol; IAM, iodoacetamide; DTT, dithiothreitol; MS, mass spectrometry; ER, endoplasmic reticulum

■

REFERENCES

(1) Baillie, T. A. (2008) Metabolism and toxicity of drugs. Two decades of progress in industrial drug metabolism. Chem. Res. Toxicol. 21, 129−137. (2) Park, B. K., Boobis, A., Clarke, S., Goldring, C. E. P., Jones, D., Kenna, J. G., Lambert, C., Laverty, H. G., Naisbitt, D. J., Nelson, S., Nicoll-Griffith, D. A., Obach, R. S., Routledge, P., Smith, D. A., Tweedie, D. J., Vermeulen, N., Williams, D. P., Wilson, I. D., and Baillie, T. A. (2011) Managing the challenge of chemically reactive metabolites in drug development. Nat. Rev. Drug Discovery 10, 292− 306. (3) Obach, R. S., Kalgutkar, A. S., Soglia, J. R., and Zhao, S. X. (2008) Can in vitro metabolism-dependent covalent binding data in liver microsomes distinguish hepatotoxic from nonhepatotoxic drugs? An analysis of 18 drugs with consideration of intrinsic clearance and daily dose. Chem. Res. Toxicol. 21, 1814−1822. (4) Bauman, J. N., Kelly, J. M., Tripathy, S., Zhao, S. X., Lam, W. W., Kalgutkar, A. S., and Obach, R. S. (2009) Can in vitro metabolismdependent covalent binding data distinguish hepatotoxic from nonhepatotoxic drugs? An analysis using human hepatocytes and liver S-9 fraction. Chem. Res. Toxicol. 22, 332−340. (5) Orr, S. T. M., Ripp, S. L., Ballard, T. E., Henderson, J. L., Scott, D. O., Obach, R. S., Sun, H., and Kalgutkar, A. S. (2012) Mechanismbased inactivation (MBI) of cytochrome P450 enzymes: Structureactivity relationships and discovery strategies to mitigate drug-drug interaction risks. J. Med. Chem. 55, 4896−4933. (6) Sakatis, M. Z., Reese, M. J., Harrell, A. W., Taylor, M. A., Baines, I. A., Chen, L., Bloomer, J. C., Yang, E. Y., Ellens, H. M., Ambroso, J. L., Lovatt, C. A., Ayrton, A. D., and Clarke, S. E. (2012) Preclinical strategy to reduce clinical hepatotoxicity using in vitro bioactivation data for >200 compounds. Chem. Res. Toxicol. 25, 2067−2082. (7) Thompson, R. A., Isin, E. M., Li, Y., Weidolf, L., Page, K., Wilson, I., Swallow, S., Middleton, B., Stahl, S., Foster, A. J., Dolgos, H.,

■

CONCLUSIONS We have developed a shotgun proteomic workflow for the characterization of human liver microsomal subproteome consisting of membrane-bound proteins. Utilizing a combination of different MS-compatible solubilizing solvents and analysis of the samples using UHPLC-MS/MS, we were able to identify a total of 2624 constituent peptides and 329 proteins in human liver microsomes with over 20% average sequence converge, which is the first report on the characterization of a human liver microsomal subproteome. This workflow was then successfully applied to identify peptides and their parent proteins adducted by a reactive drug metabolite generated in human liver microsomal incubations. A total of 18 modified proteins/19 modified constituent peptides were identified in human liver microsomes incubated directly with the reactive 2(methylsulfonyl)benzo[d]thiazole metabolite. Among these, 3 modified protein/5 modified constituent peptides were identified in incubation with parent compound 2-(methylthio)benzo[d]thiazole. This workflow does not require radiolabeled compounds and can be applied broadly in the field of drug metabolism and drug safety to understand the nature of modified proteins and precisely map the modified residues on the proteins. In the near future, we will investigate the suitability of this approach for the identification of modified peptides and proteins in hepatocytes. This label-free workflow is unbiased and generally suitable for identification and characterization of proteins adducted with reactive metabolites 1545



Article

Weaver, R., and Kenna, J. G. (2012) In vitro approach to assess the potential for risk of idiosyncratic adverse reactions caused by candidate drugs. Chem. Res. Toxicol. 25, 1616−1632. (8) Liebler, D. C. (2008) Protein damage by reactive electrophiles: Targets and consequences. Chem. Res. Toxicol. 21, 117−128. (9) Hanzlik, R. P., Fang, J., and Koen, Y. M. (2009) Filling and mining the reactive metabolite target protein database. Chem-Biol. Interact. 179, 38−44. (10) Törnqvist, M., Fred, C., Haglund, J., Helleberg, H., Paulsson, B., and Rydberg, P. (2002) Protein adducts: Quantitative and qualitative aspects of their formation, analysis and applications. J. Chromatogr. B 778, 279−308. (11) Stamper, B. D., Mohar, I., Kavanagh, T. J., and Nelson, S. D. (2011) Proteomic analysis of acetaminophen-induced changes in mitochondrial protein expression using spectral counting. Chem. Res. Toxicol. 24, 549−558. (12) Evans, D. C., Watt, A. P., Nicoll-Griffith, D. A., and Baillie, T. A. (2004) Drug-protein adducts: an industry perspective on minimizing the potential for drug bioactivation in drug discovery and development. Chem. Res. Toxicol. 17, 3−16. (13) Zhou, S. (2003) Separation and detection methods for covalent drug-protein adducts. J. Chromatogr., B 797, 63−90. (14) Boelsterli, U. A. (1993) Specific targets of covalent drug-protein interactions in hepatocytes and their toxicological significance in druginduced liver injury. Drug Metab. Rev. 25, 395−451. (15) Washburn, M. P., Wolters, D., and Yates, J. R. (2001) Largescale analysis of the yeast proteome by multidimensional protein identification technology. Nat. Biotechnol. 19, 242−247. (16) Zhang, Y., Fonslow, B. R., Shan, B., Baek, M. C., and Yates, J. R., III (2013) Protein analysis by shotgun/bottom-up proteomics. Chem. Rev. 113, 2343−2394. (17) Jeong, J., Zhang, T., Zhang, J., and Kao, Y. H. (2011) Ultra-highpressure LC (UHPLC) for therapeutic protein characterization. Am. Pharmaceut. Rev. 14, 44−51. (18) De Vos, J., Stassen, C., Vaast, A., Desmet, G., and Eeltink, S. (2012) High-resolution separations of tryptic digest mixtures using core-shell particulate columns operated at 1200bar. J. Chromatogr., A 1264, 57−62. (19) Percy, A. J., Chambers, A. G., Yang, J., Domanski, D., and Borchers, C. H. (2012) Comparison of standard- and nano-flow liquid chromatography platforms for MRM-based quantitation of putative plasma biomarker proteins. Anal. Bioanal. Chem. 404, 1089−1101. (20) Damsten, M. C., Commandeur, J. N. M., Fidder, A., Hulst, A. G., Touw, D., Noort, D., and Vermeulen, N. P. E. (2007) Liquid chromatography/tandem mass spectrometry detection of covalent binding of acetaminophen to human serum albumin. Drug Metab. Dispos. 35, 1408−1417. (21) Jenkins, R. E., Kitteringham, N. R., Goldring, C. E. P., Dowdall, S. M. J., Hamlett, J., Lane, C. S., Boerma, J. S., Vermeulen, N. P. E., and Park, B. K. (2008) Glutathione-S-transferase pi as a model protein for the characterisation of chemically reactive metabolites. Proteomics 8, 301−315. (22) Yukinaga, H., Iwabuchi, H., Okazaki, O., and Izumi, T. (2012) Glutathione S-transferase pi trapping method for generation and characterization of drug-protein adducts in human liver microsomes using liquid chromatography-tandem mass spectrometry. J. Pharmaceut. Biomed. 67−68, 186−192. (23) Pham, N. T., Jewell, W. T., Morin, D., and Buckpitt, A. R. (2012) Analysis of naphthalene adduct binding sites in model proteins by tandem mass spectrometry. Chem-Biol. Interact. 199, 120−128. (24) Kent, U. M., Sridar, C., Spahlinger, G., and Hollenberg, P. F. (2008) Modification of serine 360 by a reactive intermediate of 17-aethynylestradiol results in mechanism-based inactivation of cytochrome P450s 2B1 and 2B6. Chem. Res. Toxicol. 21, 1956−1963. (25) Lin, H. L., Zhang, H., Jushchyshyn, M., and Hollenberg, P. F. (2010) Covalent modification of Thr302 in cytochrome P450 2B1 by the mechanism-based inactivator 4-tert-butylphenylacetylene. J. Pharmacol. Exp. Ther. 333, 663−669.

(26) Zhang, H., Amunugama, H., Ney, S., Cooper, N., and Hollenberg, P. F. (2011) Mechanism-based inactivation of human cytochrome P450 2B6 by clopidogrel: Involvement of both covalent modification of cysteinyl residue 475 and loss of heme. Mol. Pharmacol. 80, 839−847. (27) Lin, H. L., Kenaan, C., and Hollenberg, P. F. (2012) Identification of the residue in human CYP3A4 that is covalently modified by bergamottin and the reactive intermediate that contributes to the grapefruit juice effect. Drug Metab. Dispos. 40, 998−1006. (28) Tzouros, M., and Pähler, A. (2009) A targeted proteomics approach to the identification of peptides modified by reactive metabolites. Chem. Res. Toxicol. 22, 853−862. (29) Blonder, J., Goshe, M. B., Moore, R. J., Pasa-Tolic, L., Masselon, C. D., Lipton, M. S., and Smith, R. D. (2002) Enrichment of integral membrane proteins for proteomic analysis using liquid chromatography-tandem mass spectrometry. J. Proteome Res. 1, 351−360. (30) Gilmore, J. M., and Washburn, M. P. (2010) Advances in shotgun proteomics and the analysis of membrane proteomes. J. Proteomics 73, 2078−2091. (31) Yang, Y., Qiu, F., Josephs, J. L., Humphreys, W. G., and Shu, Y. Z. (2012) Bioactivation of 2-(alkylthio)-1,3,4-thiadiazoles and 2(alkylthio)-1,3- benzothiazoles. Chem. Res. Toxicol. 25, 2770−2779. (32) Keller, A., Nesvizhskii, A. I., Kolker, E., and Aebersold, R. (2002) Empirical statistical model to estimate the accuracy of peptide identifications made by MS/MS and database search. Anal. Chem. 74, 5383−5392. (33) Nesvizhskii, A. I., Keller, A., Kolker, E., and Aebersold, R. (2003) A statistical model for identifying proteins by tandem mass spectrometry. Anal. Chem. 75, 4646−4658. (34) Zhang, H., Lin, Q., Ponnusamy, S., Kothandaraman, N., Teck, K. L., Zhao, C., Hon, S. K., Arijit, B., Rauff, M., Hew, C. L., Chung, M. C. M., Joshi, S. B., and Choolani, M. (2007) Differential recovery of membrane proteins after extraction by aqueous methanol and trifluoroethanol. Proteomics 7, 1654−1663. (35) Rodrigues, A. D. (1999) Integrated cytochrome P450 reaction phenotyping. Attempting to bridge the gap between cDNA-expressed cytochromes P450 and native human liver microsomes. Biochem. Pharmacol. 57, 465−480. (36) Graham, J. M. (1993) Fractionation of a microsomal fraction from rat liver or cultured cells. Methods Mol. Biol. (Clifton, NJ, U.S.) 19, 51−57. (37) Mathias, R. A., Chen, Y. S., Kapp, E. A., Greening, D. W., Mathivanan, S., and Simpson, R. J. (2011) Triton X-114 phase separation in the isolation and purification of mouse liver microsomal membrane proteins. Methods 54, 396−406. (38) Mbeunkui, F., and Goshe, M. B. (2011) Investigation of solubilization and digestion methods for microsomal membrane proteome analysis using data-independent LC-MS E. Anal. Chem. 80, 9714−9719. (39) Plazas-Mayorca, M. D., Zee, B. M., Young, N. L., Fingerman, I. M., LeRoy, G., Briggs, S. D., and Garcia, B. A. (2009) One-pot shotgun quantitative mass spectrometry characterization of histones. J. Proteome Res. 8, 5367−5374.

1546


Identification of Human Liver Microsomal Proteins Adducted by a

Recommend Documents