Artifactual Nitration Controlled Measurement of Protein-Bound 3

OCTOBER 2002 VOLUME 15, NUMBER 10 © Copyright 2002 by the American Chemical Society

Articles Artifactual Nitration Controlled Measurement of Protein-Bound 3-Nitro-L-tyrosine in Biological Fluids and Tissues by Isotope Dilution Liquid Chromatography Electrospray Ionization Tandem Mass Spectrometry Thierry Delatour,* Janique Richoz, Jacques Vuichoud, and Richard H. Stadler Nestle´ Research Center, Nestec Ltd., Vers-chez-les-Blanc, 1000 Lausanne 26, Switzerland Received May 28, 2002

A sensitive and selective method is presented to accurately determine the level of proteinbound 3-nitro-L-tyrosine (NTyr) in rat plasma and kidney samples. This assay is based on isotope dilution liquid chromatography coupled with electrospray ionization tandem mass spectrometry (LC-ESI-MS/MS). The sample preparation entails protein precipitation, acid hydrolysis with 6 N HCl, and solid-phase extraction (using reverse and aminopropyl phase cartridges) prior to the determinative step. For kidney samples, NTyr is converted into its butyl ester to improve sensitivity. The potential formation of artifactual NTyr during the acid hydrolysis step was carefully followed and determined by supplementation of the samples with 13C-labeled L-tyrosine (Tyr) prior to protein digestion. Hence, the concomitant measurement of formation of 13C-enriched NTyr enabled the accurate determination of artifactual NTyr. This approach was employed to measure the basal level of protein-bound NTyr in rat plasma and kidney samples, revealing levels in the range of 4-18 µmol/mol of Tyr and 50-68 µmol/ mol of Tyr, respectively. No artifactual nitration of Tyr was observed in kidney proteins, whereas in the case of plasma the contribution of the artifactual response ranged from 16 to 40%. This method allows the analysis of protein-bound NTyr with a full control of the artifactual nitration of tyrosine during the proteolysis and/or sample preparation. Reliable detection of NTyr in proteins may allow insight into the role of nitric oxide-derived oxidants under various pathological conditions.

Introduction Nitric oxide (•NO) is a free radical that is generated endogenously by a variety of mammalian cells, including * To whom correspondence should be addressed. E-mail: [email protected]. Phone: (+)21/785.92.20. Fax: (+)21/785.85.53.

vascular endothelium, neurons, smooth muscle cells, macrophages, neutrophils, platelets, and pulmonary epithelium (1). The physiological actions of •NO include mediating vasodilation, neurotransmission, inhibition of platelet adherence and aggregation, and the macrophage and neutrophil killing of pathogens (2). It has been shown that, in an oxygenated environment, •NO may directly react with the superoxide anion radical (O2•-), producing

10.1021/tx0200414 CCC: $22.00 © 2002 American Chemical Society Published on Web 10/04/2002

1210

Chem. Res. Toxicol., Vol. 15, No. 10, 2002

peroxynitrite (ONOO-) at almost diffusion-limited rates (3). One important pathway of peroxynitrite decomposition, catalyzed by transition metals, is the production of the hydroxyl anion (HO-) and nitronium cation (NO2+) by heterolytic cleavage. Since the electrophilic nitronium species is able to react with phenolic compounds, the sitespecific nitration of protein-bound tyrosine may procure 3-nitro-L-tyrosine (NTyr),1 a stable endproduct that is frequently employed as a diagnostic marker of reactive nitrogen species such as peroxynitrite (4). Several experiments also gave some evidence for the formation of nitric oxide (•NO) and hydroxyl (HO•) radicals as decomposition products of peroxynitrite at physiological pH (5). However, recent results suggest that peroxynitrite does not contribute considerably to the nitration of tyrosine residues in vivo (6), and other processes, that may give rise to NTyr, have been identified. For instance, it was shown that the scavenging of superoxide anion radical, inhibiting the release of peroxynitrite, had a limited effect on the level of both nitrite and NTyr in activated mouse peritoneal macrophages (7). Interestingly, the results of some experiments performed on models of acute inflammation in mice demonstrated that myeloperoxidase was involved in the formation of NTyr (8). Therefore, much effort has been devoted to the development of methods to determine NTyr in biological fluids and tissues, including immunohistochemical (9, 10) and high-performance liquid chromatography (HPLC) techniques, the latter coupled with either ultraviolet (11), electrochemical (12, 13), fluorescence (14), or mass spectrometry detection (15, 16). Analytical techniques to detect NTyr based on gas chromatography-mass spectrometry (GC-MS) (17, 18) and liquid chromatography coupled to mass spectrometry (LC-MS) (19) have rapidly evolved over the past few years, providing an improved accuracy and sensitivity. However, it was recently demonstrated that GC-MS assays may have a drawback that is associated with the propensity of certain derivatization reagents to induce the nitration of tyrosine in the presence of endogenous nitrate in the samples (20). In that study, the authors investigated the use of milder derivatization conditions and different chemicals to derivatize the target analyte, but could not fully avoid artifact formation. The utilization of LC-MS techniques circumvents, within the aspect of artifact formation, the critical derivatization step, and recently reports have been published to determine NTyr in different biological matrices (19, 20). Some GC-MS methods with a control of the artifactual formation of NTyr were also described recently (15, 21). The inconsistency of data published in the literature indicates that in some assays the artifactual nitration of Tyr may contribute significantly to the level of measured NTyr. For instance, the amount of protein-bound NTyr in rat heart was in the range of 110160 µmol/mol of tyrosine employing GC-MS (18, 22), whereas the utilization of an LC-MS/MS technique showed a level below 15 µmol/mol of tyrosine (20). Similar discrepancies were observed when NTyr in rat liver proteins was measured by either GC-MS (320-490 µmol/ mol of tyrosine) (22) or HPLC-ECD (9.5 ( 1.1 µmol/mol of tyrosine) (23). 1 Abbreviations: BSA, bovine serum albumine; LC-ESI-MS/MS, liquid chromatoghraphy coupled with electrospray ionization tandem mass spectrometry; NTyr, 3-nitro-L-tyrosine; [d3]-NTyr, 2,5,6-d3-3nitro-L-tyrosine; [13C9]-NTyr, R,β,γ,1,2,3,4,5,6-13C9-3-nitro-L-tyrosine; Tyr, L-tyrosine; [13C9]-Tyr, R,β,γ,1,2,3,4,5,6-13C9-L-tyrosine.

Delatour et al.

In this study, we focus on the potential artifactual formation of NTyr during the protein hydrolysis step. The method presented is based on LC-MS/MS and enables the accurate quantification of protein-bound NTyr. The usage of stable isotope labeled Tyr and NTyr permits the detection and quantification of the artifactual nitration of Tyr during protein digestion. To achieve adequate sensitivity in kidney tissues, an additional sample treatment step (butylation) is introduced to determine NTyr at basal levels.

Experimental Section Chemicals and Reagents. Caution: tetranitromethane is a cancer suspect product, and highly explosive in the presence of impurities. All solvents and reagents were of analytical grade. Methanol, acetonitrile, acetic acid, ammonium acetate, or formate and hydrochloric acid 37% were obtained from Merck (Darmstadt, Germany). Hydrochloric acid (3 N) in n-butanol was purchased from Regis Technologies (Morton Grove, IL), and Tyr was from Sigma-Aldrich (Steinheim, Switzerland). NTyr was obtained from Sigma (St. Louis, MO). Stable isotope enriched 2,3,5,6-d4-L-tyrosine (99.1 atom % D) and R,β,γ,1,2,3,5,6-13C9L-tyrosine ([13C9]-Tyr, 97-98 atom % 13C) were purchased from Cambridge Isotope Laboratories (Andover, MA) (for the labeling of the positions, refer to ref 24). The preparation of isotopelabeled derivatives of NTyr was based on the very high reactivity of tetranitromethane (Aldrich, Milwaukee, WI) toward Tyr (25), and synthesized as described elsewhere (24). The contribution of nonlabeled Tyr and NTyr in the synthesized standards of [13C9]-Tyr and [d3]-NTyr, measured by LC-MS, was determined as 0.27 and 0.89%, respectively Animals. Measurements of protein-bound NTyr were performed on plasma and tissues of male Fischer-344 rats (IffaCredo, L’Arbresle, France), weighing between 150 and 200 g. The animals were acclimatized for 24 h in wood-shaved cages (Nafag 890 food and water ad libitum), and then sacrificed under anesthesia (pentobarbital 60 mg/kg i.p.) to collect blood and kidneys. Heparinized blood samples were immediately centrifuged, and plasma and kidney samples stored at -80°C until analysis. Amino Acids Composition in Bovine Serum Albumin. The composition of amino acids in bovine serum albumine (BSA) was performed with a High Performance Analyzer System 6300 from Beckman Coulter (Fullerton, CA). Amino acids were separated on a cation-exchange column and detection was carried out by postcolumn derivatization with ninhydrin. Derivatized amino acids were detected spectrophotometrically at 570 and 440 nm. Preparation of Proteins from Biological Matrices. Proteins were quantified using the Coomassie Plus Protein Assay Reagent Kit from Pierce (Rockford, IL). Plasma: Cold acetone/methanol (3 mL, 50/50, v/v) was added to 500 µL of plasma. The resulting suspensions were stored for 90 min at 4 °C prior to centrifugation (Mistral 2000R apparatus, Oberwil, Switzerland) for 15 min at 550g (4 °C). The supernatants were discarded and a further 2 mL of the solvent mix acetone/ methanol (50/50, v/v) added to the protein pellets. The proteins were resuspended and again centrifuged. The supernatants were eliminated and protein pellets dissolved in water prior to quantification. About 400 µg of plasma protein was supplemented with [13C9]-Tyr (52 µg) prior to solvent evaporation under a gentle stream of nitrogen. Samples were reconstitued in 2 mL of HCl (6 N), kept for 24 h at 110 °C, and then evaporated to dryness under a stream of nitrogen. The hydrolysates were then subjected to solid-phase extraction as described below. Kidney: Rat kidney samples were pulverized under liquid nitrogen. Kidney powder was suspended in KCl (1.14%, v/v, approximately 3 mL/g of tissue) and homogenized (Braun homogenizer, Nelsungen, Germany). The resulting suspensions were centrifuged for 30 min at 10000g (4 °C). The supernatants were collected and ultracentrifuged (105000g) for 60 min in

Protein-Bound 3-Nitrotyrosine in Biological Matrices

Chem. Res. Toxicol., Vol. 15, No. 10, 2002 1211

Table 1. Mass Transitions Used for the Detection of NTyr, Tyr, and Their Isotope-Labeled Derivatives by LC-ESI-MS/MS in Rat Plasma and Kidney Proteins transition (m/z precursor ion f m/z product ion)

3-nitro-L-tyrosine 2,5,6-d3-3-nitro-L-tyrosine R,β,γ,1,2,3,4,5,6-13C9-3-nitro-L-tyrosine

quantification transition (m/z)

collision energy (eV)

confirmation transition (m/z)

collision energy (eV)

227 f 181 283 f 181 230 f 184 286 f 184 236 f 189 292 f 189 182 f 165 191 f 174

20 25 20 25 20 25 22 22

259 f 181 283 f 227 262 f 184 286 f 230 268 f 189 292 f 236 182 f 136 191 f 144

24 15 24 15 24 15 28 28

nativea butyl esterb nativea butyl esterb nativea butyl esterb

L-tyrosine c

R,β,γ,1,2,3,4,5,6-13C9-L-tyrosinec

a Quantif. trans./confirm. trans. ) 7.9 ( 0.4. b Quantif. trans./confirm. trans. ) 6.5 ( 0.3. 0.01.

order to isolate the cytosol (supernatant). Subsequently, 7 mL of cold methanol/acetone (50/50, v/v) was added to about 2 mg of cytosolic proteins, and the samples kept for 2 h at 4 °C. The samples were then centrifuged for 10 min at 550g (4 °C). The supernatants were eliminated, and protein pellets spiked with [13C9]-Tyr (65 µg) prior to protein digestion, which was conducted as described above for the plasma samples. The hydrolysates were evaporated to dryness under a stream of nitrogen prior to solid-phase extraction. Solid-Phase Extraction. Dry sample extracts of plasma and kidney were spiked with [d3]-NTyr (1.10 and 3.66 ng, respectively) and reconstitued in 1 mL of ammonium acetate (2 mM, pH 4.35). The extracts were loaded onto preconditioned C-18 Bond Elut cartridges (500 mg, Varian, Middelburg, The Netherlands). The columns were subsequently washed with 1 mL of ammonium acetate 2 mM, pH 4.35, and Tyr was eluted with 1 mL of methanol/ammonium acetate 2 mM, pH 4.35 (5/95 v/v). The columns were then washed with 1 mL of methanol/ ammonium acetate 2 mM, pH 4.35 (10/90 v/v), and NTyr was eluted with 1.5 mL of methanol/ammonium acetate 2 mM, pH 4.35 (20/80 v/v). The NTyr-containing fractions were submitted to a second purification step on aminopropyl Isolute-cartridges (500 mg, Separtis, Grellingen, Switzerland). Fractions were applied onto preconditionned columns and washed with 2 mL of ammonium formate (2 mM, pH 3.55). Elution of NTyr was achieved with 1.5 mL of water/acetic acid 25/15 (v/v). The eluates were concentrated to dryness in vacuo and kept at -20 °C until analysis. Derivatization of NTyr. After solid-phase extraction, the dry fractions of the kidney samples were resuspended in 2 mL of HCl (3 N) in n-butanol, and sonified for approximately 10 min to achieve a homogeneous suspension. Derivatization was performed for 1 h at 65 °C, and thereafter the reaction mixture evaporated to dryness in vacuo at 25 °C. Liquid Chromatography-Electrospray Ionization Tandem Mass Spectrometry (LC-ESI-MS/MS). The fractions intended for measurement of NTyr were reconstitued in 30 µL of water, whereas 600 µL of water were used to reconstitute fractions for the determination of Tyr. The HPLC system consisted of a HP series 1100 (Hewlett-Packard, Palo-Alto, CA) and a XTerra MS C18 microbore column (150 × 1.0 mm i.d. 3.5 µm, Waters, Milford, MA), operated at a flow rate of 50 µL/min. Injection volume: 10 µL. The separation of NTyr and Tyr was achieved with Gradient 1, whereas Gradient 2 was used to determine NTyr butyl ester (only in kidney tissue). Gradient 1: 0-5 min, 100% of water/formic acid 1000/0.8 (v/v) (solvent A); 5-19 min, linear ramp to 63% of methanol/water/formic acid 1000/250/0.2 (v/v/v) (solvent B); 19-20 min, linear ramp to 100% solvent B (during 4 min). Gradient 2: 0-1 min, 100% of water/ formic acid 1000/0.8 (v/v) (solvent A); 1-10 min, linear ramp to 100% of methanol/formic acid 1000/0.4 (v/v) (solvent B); 10-25 min, 100% of solvent B. Typical retention times: NTyr, 17.9 min (gradient 1); NTyr butyl ester, 19.3 min (gradient 2); Tyr, 6.7 min. Detection was performed by positive electrospray ionization tandem mass spectrometry using a Finnigan MAT TSQ 7000 equipped with the API 2 interface (San Jose, CA).

c

Quantif. trans./confirm. trans. ) 0.40 (

MS parameters were as follows: spray voltage: 1.1 kV; capillary temperature, 360 °C; sheath gas pressure, 90 psi; peak width, 1.0 Da; scan time per transition, 500 ms; collision gas pressure, 2.8 mTorr. For added certainty of analyte identity, the acquisition was carried out in the selected reaction monitoring mode choosing two characteristic mass transitions per analyte (Table 1). Quantification. Calibration curves (5-point) for the quantification of NTyr were constructed over the range 70-4500 pg and 10-1800 pg for kidney tissues and plasma, respectively. Each calibrator was supplemented with [d3]-NTyr and [13C9]Tyr (containing [13C9]-NTyr as an impurity at a level of 22.4 µmol/mol of [13C9]-Tyr). [d3]-NTyr was employed as an internal standard to accurately measure NTyr and [13C9]-NTyr. Tyr was determined by establishing a calibration curve spanning from 30 to 960 ng and adding [13C9]-Tyr as internal standard. For quantitation, the area ratios (see quantification transitions in Table 1) were plotted against the NTyr or Tyr amount. A linear regression model was used to estimate the linearities. Inter- and intra-assay precisions were determinated according to Miller and Miller (26).

Results Fragmentation by Collision-Induced Dissociation. Flow injection analysis was conducted on the standard compounds of Tyr, NTyr, and NTyr butyl ester. The assignments of fragment ion structures were assisted by the analysis of isotope-labeled compounds. Fragmentation of the NTyr protonated molecular ion at m/z 227 gives rise to an intense fragment ion at m/z 181, corresponding to the loss of HCOOH. It has been suggested that the ion observed at m/z 181 corresponds to a loss of the nitro group (19). However, this structure does not match with the results we obtained with isotope-labeled compounds. In fact, the structure suggested by Althaus et al. (19) should procure an ion at m/z 186 for NTyr solubilized in D2O, which differs from m/z 184 observed in our spectrum. Furthermore, the ionization of NTyr in the presence of methanol generates an ion at m/z 259, which is assigned to a protonated methanol adduct. The mass difference of ions at m/z 259 and 227 (∆M ) 32 Da) corresponds to the molecular mass of methanol, substantiated by the lack of m/z 259 when replacing methanol with acetonitrile. As briefly described elsewhere (19), the protonated molecular ion of NTyr butyl ester ([M + H]+ at m/z 283) shows an elimination of HCOO(CH2)3CH3 and C4H8, producing fragment ions at m/z 181 and 227, respectively. In the case of Tyr, the loss of HCOOH is again observed by collision-induced dissociation. The fragment ion spectrum of the pseudo-molecular ion at m/z 182 exhibits an intense fragment at m/z 136, corresponding to a mass loss of 46 Da. In addition, the fragmentation

1212


Delatour et al.

Figure 1. Selected reaction monitoring chromatogram of a rat plasma protein sample containing Tyr at 416 nmol/mg of protein and NTyr at 14.5 µmol/mol of Tyr (394 µg of protein was fortified with 51.8 µg of [13C9]-Tyr and the hydrolysate with 1.10 ng of [d3]-NTyr).

pattern exhibits an ion at m/z 165, which indicates a loss of NH3 (∆M ) 17 Da). Proteolysis and Stability of NTyr in Hydrochloric Acid. An enzymatic hydrolysis may be considered as the initial method of choice to avoid any potential artifactual nitration of Tyr. While an enzymatic hydrolysis is often specific and rapid, it is mostly never complete, and a significant part of the original protein (2-20%) may remain undigested (27). Consequently, the NTyr level may be underestimated if enzyme-resistant peptides harbor a disproportionated amount of NTyr residues. To avoid this drawback, acid hydrolysis was assessed, and proteins were cleaved with hydrochloric acid (6 N) to ensure an efficient release of the amino acids (28). To validate our experimental conditions, a standard of BSA was subjected to acid hydrolysis over a period of 24 h. The amino acid composition of the resulting hydrolysate was evaluated and, for each amino acid, the calculated amount was compared to the expected values (29, 30). For 13 of 20 detected amino acids, the relative expected value (calculated amount vs expected) ranged from 90 to 103%. This indicates that our acid hydrolysis conditions are suitable, ensuring a good release of amino acids. The completeness of the cleavage was confirmed by the relative expected values of slowly released amino acids such as Val, Ile, and Leu at 95, 89, and 97%, respectively (31). The well-known losses of Ser, Cys, Met, and Trp (32) were confirmed (relative expected value below 90%), and the conversion of Asn and Gln (not detected) into Asp and Glu (relative expected value at 128%), respectively, was also observed (32). Acid hydrolysis is relatively harsh and thus a further potential concern was the stability of both NTyr and Tyr.

Table 2. Stability of NTyr and Tyr in Hydrochloric Acid 6 N at 110 °C (n ) 3, mean ( SD). incubation time nominal amount

24 h

48 h

3-nitro-L-tyrosine 2.08 pmol 1.9 ( 0.2 pmol 1.91 ( 0.06 pmol L-tyrosine 60.8 nmol 58.0 ( 3.3 nmol 47.0 ( 2.8 nmol

Thus, standards of NTyr (470 pg) and Tyr (11 µg) were subjected to a 24 and 48 h incubation in 6 N HCl at 110 °C (n ) 3). The samples were subsequently evaporated to dryness, spiked with their corresponding internal standards, and analyzed by LC-ESI-MS/MS. The results demonstrate that the degradation of both NTyr and Tyr over a period of 24 h does not differ significantly (R ) 0.05) (Table 2). Prolonged storage in 6 N HCl (48 h) revealed a loss of approximately 9 and 22% of the initial amount of NTyr and Tyr, respectively. Therefore, proteinbound NTyr can be expressed relative to Tyr without the introduction of a correction factor. Method Performance Characteristics. The parameters of the electrospray ion source were optimized to obtain the best response for NTyr, setting a target for the limit of detection of the analyte at the low picogram level (on-column). Notably, the same conditions were employed for the detection of Tyr, even though the relative amount of Tyr in the samples was approximately 4-5 orders of magnitude greater than that of NTyr. Greater certainty of analyte identity was achieved by monitoring two characteristic transitions per analyte. Plasma: The proteolysis of 400 µg (133 µg equivalent injected) of plasma protein provided a response of NTyr with a signal-to-noise ratio ranging from 8 to 25 together with a percentage of expected value (ratio [d3]-NTyr



Figure 2. Selected reaction monitoring chromatograms of standards of either NTyr or its butyl ester (290 fmol injected).

response in a sample/[d3]-NTyr response in a standard) measured at 45.5 ( 0.1%. In addition, the coefficient of variation (CV) of the [d3]-NTyr response within samples (n ) 6) was 9.8%. These results suggest that 400 µg is a suitable amount of protein to detect NTyr at the fmol level (Figure 1) with a good precision. In addition, the intra- and interassay precisions were in the range 3.423% (mean: 10%) and 14-26% (mean: 21%), respectively. Kidney: In the case of kidney tissues (1.95 mg of protein, 650 µg injected), the signal-to-noise ratio of endogenous NTyr was too low to allow an accurate quantitation of the analyte. The four samples that were analyzed provided signal-to-noise ratios of NTyr ranging from 2.9 to 11.1 (7.0 ( 2.7, mean ( SD). The typical background level of the transition m/z 227 f 181 was approximately 5-fold lower in plasma samples than in kidney tissues; the higher background level observed in tissues may explain the loss of NTyr sensitivity in this matrix. Dilution of the extract also indicated that the poor NTyr response may be due, at least in part, to some ion suppression effect. However, adequate analyte sensitivity could not be achieved, and thus we chose to react NTyr to its butyl ester (33). This step afforded a 7-fold increase in the signal-to-noise ratio versus the native analyte, based upon the mass transitions of the standard compounds (Figure 2). Analysis of tissue samples (n ) 4) revealed a signal-to-noise ratio ranging from 20 to 50, with a mean ( SD at 36 ( 13. This represents a 5-fold increase with respect to the nonderivatized NTyr. In the case of derivatized samples, the CV of the peak area of the internal standard ([d3]-NTyr) was calculated at 11.9% (based on four kidney samples). In comparison, the CV was 18.1% when using the same set of samples analyzed before the derivatization step. Thus, the precision of the method is significantly improved by derivatizing NTyr to its butyl ester prior to the determinative step. Typical MS chromatograms of a rat kidney sample are depicted in Figure 3. Specificity and Accuracy of the Method. To ascertain the level of NTyr in kidney proteins, we investigated the accuracy and the selectivity of the method. After the analysis of the extracts of samples A and B (Table 3), the response ratios NTyr butyl ester/[d3]-NTyr butyl ester were 1.2 and 2.0, respectively. Then, these extracts were spiked with a mixture containing standards

Table 3. Comparison of the Response Ratio NTyr Butyl Ester/[d3]-NTyr Butyl Ester Measured in Two Rat Kidney Samples after Addition of Butyl Ester Standards in the Extracts sample

extract

standards additiona

relative difference (%)

Ab B

1.2 2.0

1.1 1.9

-8.3 -5.0

a Extracts were spiked with butyl esters of both 3-nitro-Ltyrosine and 2,5,6-d3-3-nitro-L-tyrosine in the same ratio as initially measured in the extract (Extract column). b Diet and tissues collection procedure not controlled.

of both NTyr and [d3]-NTyr butyl esters in the same ratio than the one measured initially in the samples. Spiked extracts were reanalyzed, and it was found that the differences did not exceeded (10%, suggesting that the sample preparation from biological tissues does not induce any significant trend. Artifactual Nitration of Tyr and Determination of Endogenous Level of NTyr in Artifact-Induced Samples. Plasma may contain significant levels of endogenous nitrite or nitrate (10-40 µM) (34), and proteolysis under strong acid conditions could lead to the nitration of Tyr residues. Therefore, the sample digestion step may introduce an error of potential overestimation. This important aspect was investigated by supplementing plasma and kidney samples (note: in the case of kidney, the experiment was performed with cytosol and not with isolated proteins) with sodium nitrite. Nitration of Tyr was induced during the proteolysis step (Figure 4) by employment of either 13.5 or 108 µg/mg protein. The results (Table 4) demonstrate that no artifactual nitration of Tyr occurred in the original (nonnitrite treated) samples. In fact, the [13C9]-NTyr levels in plasma (21.2 µmol/mol of [13C9]-Tyr) and kidney (22.1 µmol/mol of [13C9]-Tyr) were similar to those measured in the reference (reference: standard solution of [13C9]-Tyr which was not submitted to a 24 h treatment in HCl 6 N), at a level of 22.4 µmol of [13C9]-NTyr/mol of [13C9]-Tyr. In contrast, the same samples supplemented with sodium nitrite (13.5 µg/mg of protein) contained a higher level of [13C9]-NTyr, corresponding to a 3- and 1.4-fold augmentation in the case of plasma and kidney tissue, respectively. This increase is due to the artifactual

1214


Delatour et al.

Figure 3. Selected reaction monitoring chromatogram of a rat kidney proteins sample containing Tyr at 251 nmol/mg of protein and NTyr at 50.6 µmol/mol of Tyr (1.96 mg of protein was fortified with 64.8 µg of [13C9]-Tyr and the hydrolysate with 3.66 ng of [d3]-NTyr).

Figure 4. Selected reaction monitoring chromatogram of a rat kidney cytosol sample without any addition of salt (A) and spiked with sodium nitrite at 13.5 µg/mg of protein (B) and 108 µg/mg of protein (C) (924 µg of protein was fortified with 64.8 µg of [13C9]Tyr and the hydrolysate with 3.66 ng of [d3]-NTyr). Only signals used for the quantification are shown. Traces: (0) NTyr; (") [d3]NTyr; (9) [13C9]-NTyr.



Table 4. Measurements of NTyr, [13C9]-NTyr and Tyr in Original Rat Plasma Proteins and Kidney Cells Cytosols, and Same Samples Supplemented with Sodium Nitrite (at 13.5 and 108 µg/mg of protein) Prior to the Proteolysis in HCl (n ) 4)

measured NTyr (µmol/mol of Tyr)

[13C9]-NTyr (µmol/mol of [13C9]-Tyr)

artifactual [13C9]-NTyr (µmol/mol of [13C9]-Tyr)

endogenous NTyr (µmol/mol of Tyr)

364 ( 2 379 ( 1 >HLOQb

22.6 ( 1.0 69.3 ( 7.1 >HLOQc

22.4 ( 0.6 21.2 ( 0.1 69.0 ( 3.5 >HLOQc

46.6 ( 3.6 nd

22.7 ( 7.9 nd

343 ( 5 349 ( 2 525 ( 4

51.1 ( 0.6 62.0 ( 1.0 >HLOQc

22.4 ( 0.7 22.1 ( 1.5 30.9 ( 0.6 >HLOQc

8.5 ( 0.9 nd

53.5 ( 1.3 nd

Tyr (nmol/mg of protein) plasma referencea sample +nitrite 13.5 µg/mg of protein +nitrite 108 µg/mg of protein kidney referencea sample +nitrite 13.5 µg/mg of protein +nitrite 108 µg/mg of protein

a The [13C ]-NTyr level in the standard of [13C ]-Tyr was measured with five calibrators spiked with [13C ]-Tyr. b HLOQ (highest limit 9 9 9 of quantification) ) 950 nmol/mg of protein. c HLOQ ) 575 µmol/mol of Tyr.

nitration of [13C9]-Tyr, induced during the acid hydrolysis step. Therefore, the difference between the [13C9]-NTyr level in the sodium nitrite-containing samples and the nontreated samples provide an index of artifactual contribution of [13C9]-NTyr to the measurement. In the case of plasma, the artifact level was determined at 46.6 ( 3.6 µmol of [13C9]-NTyr/mol of [13C9]-Tyr, and 8.5 ( 0.9 µmol of [13C9]-NTyr/mol of [13C9]-Tyr in the case of kidney. Deduction of the artifact-induced contributions from the measured levels of NTyr affords NTyr concentrations at 22.7 ( 7.9 and 53.5 ( 1.3 µmol/mol of Tyr for plasma and kidney tissue, respectively. Moreover, these values did not differ statistically (R ) 0.05) from the ones measured in the nontreated sample (22.6 ( 1.0 and 51.1 ( 0.6 µmol of NTyr/mol of Tyr, respectively). We also observed that the calculated amount of Tyr in protein was higher in the case of sodium nitrite-treated versus nontreated samples. The kinetic of the nitritemediated nitration of Tyr in 6 N HCl is very likely dependent on the concentration of nitrite in the solution. When sodium nitrite is fortified at 13.5 µg/mg of protein (molar ratio nitrite/Tyr ) 0.54), the kinetic of the hydrochloric acid-mediated proteolysis appears to be faster than the nitration of Tyr. In this case, both Tyr and [13C9]-Tyr are soluble, and the nitration agent (sodium nitrite) can nitrate Tyr and its 13C-labeled derivative in an equivalent manner. In contrast, when sodium nitrite is added at a final concentration of 108 µg/mg of protein (molar ratio nitrite/Tyr ) 4.3), the kinetic of the nitration is faster than the release of the amino acids. Therefore, [13C9]-Tyr can be nitrated easier than Tyr which is still (at least in part) covalently bound to proteins. Consequently, the endogenous NTyr level expressed with respect to Tyr may not be accurate if the artifactual contribution is too high. To avoid this potential error, the corrected levels of NTyr should be based on the amount of protein, or Tyr should be measured independently by a suitable technique. Inhibition of Artifactual Nitration of Tyr by Phenol. It has been reported that phenolic derivatives are inhibitors of peroxynitrite-mediated nitration of Tyr (35, 36), and the phenolic group facilitates reaction of the aromatic ring with electrophiles, especially at the ortho position. Thus, phenol was added to biological samples to avoid nitrite- and/or nitrate-induced artifactual nitration of Tyr during the hydrolysis of proteins with HCl 6 N (18, 37), but no control of the efficacy of this preserva-

Table 5. Effect of Phenol on the Level of [13C9]-NTyr in [13C9]-Tyr-Spiked (737 nmol/mg of protein) Rat Plasma Proteins Treated with HCl 6 N (n ) 4) a.

referenceb

plasma

plasma + plasma + nitritec + nitritec 1% phenol

19.2 ( 0.8 22.4 ( 0.9 49.5 ( 0.4 38.2 ( 0.4 measured NTyre 24.1 ( 1.6 52.5 ( 0.9 36.9 ( 1.3 endogenous NTyre 20.9 ( 1.8 22.3 ( 1.2 17.9 ( 1.5 [13C

d 9]-NTyr

a Tyr level at 370 nmol/mg of protein. b Basal content of [13C ]9 NTyr in the standard of [13C9]-Tyr (n ) 5). c Sodium nitrite at 10.9 µg/mg of protein. d µmol per mol of [13C9]-Tyr. e µmol per mol of Tyr.

tive was carried out. We therefore investigated this aspect by measuring the level of [13C9]-NTyr in protein supplemented with [13C9]-Tyr prior to the proteolysis step. Sodium nitrite-containing rat plasma proteins (10.9 µg/ mg of protein) were hydrolyzed in the presence of phenol at a final concentration of 1% (v/v). The level of [13C9]NTyr was determined and compared to a reference (a standard of [13C9]-Tyr which was not hydrolyzed) and a control plasma (without addition of both phenol and nitrite). A slight increase in the [13C9]-NTyr level was observed in the plasma sample when compared to the reference (Table 5) and, as anticipated, the amount of [13C9]-NTyr was increased significantly (2.5-fold) when proteins were supplemented with sodium nitrite. The addition of 1% of phenol prior to hydrolysis resulted in a decrease in the artifact formation versus the nontreated control. However, this addition does not seem to completely avoid the nitration of Tyr. Comparison of the amount of both [13C9]-NTyr and NTyr measured in the nitrite + phenol-supplemented samples shows that the levels are still approximately 65% above the basal level of the control sample.

Discussion A prerequisite to the employment of a biomarker as a potential link to certain pathological and degenerative conditions is the availability of validated analytical methods that are accurate and reproducible. Especially, markers that reflect oxygen or nitrogen free radical activity in biological tissues are inherently prone to artifactual formation during the analytical procedure. Artifactual formation of the endpoint may be difficult to control, and thereby efforts should be devoted during analytical method development to address this important

1216


Delatour et al.

Table 6. Measurement of Protein-Bound NTyr in Rat Plasma and Kidney Proteins at Basal Level (n ) 4) with Quantitative Determination of the Hydrochloric Acid-Induced Artifactual Nitration of Tyr

plasma referencea 1 2 3 4 5 6 kidney reference a 1 2 3 4

measured Tyr (nmol/mg of protein)

NTyr (µmol/mol of Tyr)

[13C9]-NTyr (µmol/mol of [13C9]-Tyr)

artifactual [13C9]-NTyr (µmol/mol of [13C9]-Tyr)

endogenous NTyr (µmol/mol of Tyr)

388 ( 8 397 ( 8 416 ( 7 277 ( 2 353 ( 6 390 ( 8

21.6 ( 0.3 17.9 ( 0.3 14.5 ( 0.3 10.2 ( 0.4 8.5 ( 0.3 7.4 ( 0.2

22.4 ( 0.6 25.9 ( 0.2 27.6 ( 0.2 27.4 ( 0.2 24.4 ( 0.2 25.8 ( 0.3 25.3 ( 0.2

3.5 ( 0.6 5.2 ( 0.6 5.0 ( 0.6 2.0 ( 0.6 3.4 ( 0.7 2.9 ( 0.6

18.1 ( 0.7 12.7 ( 0.7 9.5 ( 0.7 8.2 ( 0.7 5.1 ( 0.8 4.5 ( 0.6

235 ( 8 248 ( 14 239 ( 18 238 ( 31

50.8 ( 4.1 67.3 ( 9.6 50.6 ( 6.8 54 ( 14

22.4 ( 0.7 22.8 ( 1.0 23.9 ( 1.3 24.0 ( 0.0 22.4 ( 3.2

n.s.b n.s. n.s. n.s.

a The contribution of [13C ]-NTyr in [13C ]-Tyr was measured with the five calibrators spiked with [13C ]-Tyr. b n.s.: no statistical 9 9 9 significant difference between sample and reference (R ) 0.05).

aspect. A good example of frequently overestimated background levels of biomarkers are oxidized bases of DNA (38), and in particular 8-oxo-7,8-dihydroguanine (39), which reflects oxygen free-radical induced damage to DNA (40, 41). Similar observations and comments have been raised with regard to NTyr in plasma and tissues (20), emphasizing the need of well-devised and carefully validated analytical tools in order to generate meaningful results that reflect subtle but significant differences in biomarker levels. The method developed in this study includes the measurement of [13C9]-NTyr, which compensates for any artifactual formation of the target analyte. In the case of kidney tissues, no significant increase of NTyr versus the reference was observed, concluding that the artifactual nitration of Tyr during the acid hydrolysis step is negligible. Calculation of the NTyr level in kidney proteins without applying a correction factor on the measured values showed that NTyr concentration ranged from 50 to 67 µmol/mol of Tyr, with a median of 51 µmol/ mol of Tyr. In contrast, a slight artifactual contribution was observed in plasma. Indeed, the [13C9]-NTyr level in each sample was significantly above the reference value, but the contribution of the artifact never exceeded 40% of the measured value (Table 6). Consequently, the NTyr amounts were corrected to obtain a reliable basal NTyr concentration. In rat plasma, the range was 4.8-18.4 µmol/mol of Tyr with a median at 8.9 µmol/mol of Tyr. For kidney samples, derivatization of NTyr to its corresponding butyl ester increased the sensitivity of the method, enabling accurate determination of endogenous amounts of the marker in the tissue and biological fluid samples (24). A further contributing factor to the improvement in sensitivity may be the elimination of insoluble particles present in the sample extract, resulting in more efficient butylation of the homogeneous reaction mix. To ascertain the absence of discriminate reactivity during the derivatization step, the ratio of the [d3]-NTyr response/[13C9]-NTyr response was measured before and after derivatization. Indeed, the intensities of [d3]-NTyr, [13C9]-NTyr and their respective butyl ester were adequate to ensure an accurate integration of the signals. A paired t-tested was applied to eight samples (four kidney samples prepared in duplicate) analyzed before and after derivatization. The results show that

derivatization does not induce any significant variation (R ) 0.05) of the response ratio [d3]-NTyr/[13C9]-NTyr, which corroborates that the derivatization reaction does not induce any trend of the response due to inaccessibility or artifactual nitration of residual Tyr present in NTyr extracts during the treatment. The addition of phenol (1%) in the sample decreases the artifactual nitration of Tyr, which may occur during the acid hydrolysis step, but does not lead to a complete suppression. The nitration of Tyr is a very efficient reaction under acidic conditions (42) and a competitive reaction leads to the nitration of both phenol and Tyr. Since the content of Tyr in proteins is 105-106-fold higher than the level of NTyr, even a slight nitration of Tyr induces an overestimation of the amount of NTyr residues. Thus, the use of phenol as an inhibitor of Tyr nitration does not appear to be a suitable approach to estimate the endogenous level of protein-associated NTyr when the proteolysis is carried out with hydrochloric acid. At least, its concentration should be optimized to obtain an artifactual contribution of NTyr well below the endogenous basal level. The basal level of NTyr in rat plasma has been reported previously, based on a method using HPLC coupled with a dual-mode electrochemical detector (12). The authors calculated basal levels at 23 ( 7 and 57 ( 23 µmol/mol of Tyr, depending on the diet fed to the animals. In essence, our results are consistent with these findings and confirm that the basal level of NTyr in rat plasma proteins may be subjected to significant variations. In the case of kidney protein tissues, the basal level of NTyr was found to be significantly higher than in plasma (approximately 5-fold). Interestingly, as far as our study is concerned, the interanimal variation of NTyr in rat kidney protein seems to be much lower (CV ) 14%, n ) 4) than in plasma protein samples (CV ) 52%, n ) 6). Additional studies will now be conducted to confirm these observations.

References (1) Moncada, S., and Higgs, E. A. (1991) Endogenous nitric oxide: physiology, pathology and clinical relevance. Eur. J. Clin. Invest. 21, 361-374. (2) Rubbo, H., Darley-Usmar, V., and Freeman, B. A. (1996) Nitric oxide regulation of tissue free radical injury. Chem. Res. Toxicol. 9, 809-820.

Protein-Bound 3-Nitrotyrosine in Biological Matrices (3) Huie, R. E., and Padmaja, S. (1993) Reaction of NO with O2•-. Free Radical Res. Commun. 18, 195-199. (4) Beckman, J. S., and Koppenol, W. H. (1996) Nitric oxide, superoxide, and peroxynitrite: the good, the bad, and the ugly. Am. J. Physiol. 271, C1424-C1437. (5) Halliwell, B., Zhao, K., and Whiteman, M. (1999) Nitric oxide and peroxynitrite. The ugly, the uglier and the not so good: a personal view of recent controversies. Free Radical Res. 31, 651-669. (6) Pfeiffer, S., and Mayer, B. (1998) Lack of tyrosine nitration by peroxynitrite generated at physiological pH. J. Biol. Chem. 273, 27280-27285. (7) Pfeiffer, S., Lass, A., Schmidt, K., and Mayer, B. (2001) Protein tyrosine nitration in mouse peritoneal macrophages activated in vitro and in vivo: evidence against an essential role of peroxynitrite. FASEB J. 15, 2355-2364. (8) Gaut, J. P., Byun, J., Tran, H. D., Lauber, W. M., Carroll, J. A., Hotchkiss, R. S., Belaaouaj, A., and Heinecke, J. W. (2002) Myeloperoxidase produces nitrating oxidants in vivo. J. Clin. Invest. 109, 1311-1319. (9) Miller, M. J. S., Thompson, J. H., Zhang, X.-J., SadowskaKrowicka, H., Kakkis, J. L., Munshi, U. K., Sandoval, M., Rossi, J. L., Eloby-Childress, S., Beckman, J. S., Ye, Y. Z., Rodi, C. P., Manning, P. T., Currie, M. G., and Clark, D. A. (1995) Role of inducible nitric oxide synthase expression and peroxynitrite formation in guinea pig ileitis. Gastroenterology 109, 14751483. (10) Ischiropoulos, H., Beers, M. F., Ohnishi, S. T., Fisher, D., Garner, S. E., and Thom, S. R. (1996) Nitric oxide production and perivascular tyrosine nitration in brain after carbon monoxide poisoning in the rat. J. Clin. Invest. 97, 2260-2267. (11) Kaur, H., and Halliwell, B. (1994) Evidence for nitric oxidemediated oxidative damage in chronic inflammation. Nitrotyrosine in serum and synovial fluid from rheumatoid patients. FEBS Lett. 350, 9-12. (12) Sodum, R. S., Akerkar, S. A., and Fiala, E. S. (2000) Determination of 3-nitrotyrosine by high-pressure liquid chromatography with a dual-mode electrochemical detector. Anal. Biochem. 280, 278-285. (13) Hensley, K., Maidt, M. L., Yu, Z., Sang, H., Markesbery, W. R., and Floyd, R. A. (1998) Electrochemical analysis of protein nitrotyrosine and dityrosine in the Alzheimer brain indicates region-specific accumulation. J. Neurosci. 18, 8126-8132. (14) Kamisaki, Y., Wada, K., Nakamoto, K., Kishimoto, Y., Kitano, M., and Itoh, T. (1996) Sensitive determination of nitrotyrosine in human plasma by isocratic high-performance liquid chromatography. J. Chromatogr. B 685, 343-347. (15) Frost, M. T., Halliwell, B., and Moore, P. (2000) Analysis of free and protein-bound nitrotyrosine in human plasma by a gas chromatography/mass spectrometry method that avoids nitration artifacts. Biochem. J. 345, 453-458. (16) Jiao, K., Mandapati, S., Skipper, P. L., Tannenbaum, S. R., and Wishnok, J. S. (2001) Site-selective nitration of tyrosine in human albumine by peroxynitrite. Anal. Biochem. 293, 43-52. (17) Schwedhelm, E., Tsikas, D., Gutzki, F.-M., and Fro¨lich, J. C. (1999) Gas chromatography-tandem mass spectrometric quantification of free 3-nitrotyrosine in human plasma at the basal level. Anal. Biochem. 276, 195-203. (18) Crowley, J. R., Yarasheski, K., Leeuwenburgh, C., Turk, J., and Heinecke, J. W. (1998) Isotope dilution mass spectrometric quantification of 3-nitrotyrosine in proteins and tissue is facilitated by reduction to 3-aminotyrosine. Anal. Biochem. 259, 127135. (19) Althaus, J. S., Schmidt, K. R., Fountain, S. T., Tseng, M. T., Carroll, R. T., Galatsis, P., and Hall, E. D. (2001) LC-MS-MS detection of peroxynitrite-derived 3-nitrotyrosine in rat microvessels. Free Radical Biol. Med. 29, 1085-1095. (20) Yi, D., Ingelse, B. A., Duncan, M. W., and Smythe, G. A. (2000) Quantification of 3-nitrotyrosine in biological tissues and fluids: generating valid results by eliminating artifactual formation. J. Am. Soc. Mass Spectrom. 11, 578-586. (21) Gaut, J. P., Byun, J., Tran, H. D., and Heinecke, J. W. (2002) Artifact-free quantification of free 3-chlorotyrosine, 3-bromotyrosine, and 3-nitrotyrosine in human plasma by electron capturenegative chemical ionization gas chromatography mass spectrometry and liquid chromatography-electrospray ionization tandem mass spectrometry. Anal. Biochem. 300, 252-259.

Chem. Res. Toxicol., Vol. 15, No. 10, 2002 1217 (22) Leeuwenburgh, C., Hansen, P., Shaish, A., Holloszy, J. O., and Heinecke, J. W. (1998) Markers of proteins oxidation by hydroxyl radical and reactive nitrogen species in tissues of aging rats. Am. J. Physiol. 274, R453-R461. (23) Skinner, K. A., Crow, J. P., Skinner, H. B., Chandler, R. T., Thompson, J. A., and Parks, D. A. (1997) Free and proteinassociated nitrotyrosine formation following rat liver preservation and transplantation. Arch. Biochem. Biophys. 342, 282-288. (24) Delatour, T., Guy, P. A., Stadler, R. H., and Turesky, R. J. (2002) 3-Nitrotyrosine butyl ester: a novel derivative to assess tyrosine nitration in rat plasma by liquid chromatography-tandem mass spectrometry detection. Anal. Biochem. 302, 10-18. (25) Sokolowsky, M., Riordan, J. F., and Vallee, B. L. (1966) Tetranitromethane. A reagent for the nitration of tyrosyl residues in proteins. Biochemistry 5, 3582-358. (26) Miller, J. C., and Miller, J. N. (1993) Statistics for analytical chemistry, Ellis Horwood, London. (27) Shigenaga, M. K. (1999) Quantitation of protein-bound 3-nitrotyrosine by high-performance liquid chromatography with electrochemical detection. Methods Enzymol. 301, 27-40. (28) Penke, B., Ferenczi, R., and Kovacs, K. (1974) A new hydrolysis method for determining tryptophan in peptides and proteins. Anal. Biochem. 60, 45-50. (29) Brown, J. R. (1979) Structure of bovine serum albumine. Fed. Proc. 34, 591. (30) Hirayama, K., Akashi, S., Furuya, M., and Fukuhara, K.-I. (1990) Rapid confirmation and revision of the primary structure of bovine serum albumine by ESIMS and Frit-FAB-LC/MS. Biochem. Biophys. Res. Commun. 173, 639-646. (31) Kellner, R., Lottspeich, F., and Meyer, H. E. (1994) Microcharacterization of proteins, VCH Verlagsgesellschaft, Weinheim. (32) Inglis, A. S. (1983) Single hydrolysis method for all amino acids, including cysteine and tryptophan. Methods Enzymol. 91, 2636. (33) Chace, D. H., Millington, D. S., Terada, N., Kahler, S. G., Roe, C. R., and Hofman, L. F. (1993) Rapid diagnosis of phenylketonuria by quantitative analysis for phenylalanine and tyrosine in neonatal blood spots by tandem mass spectrometry. Clin. Chem. 39, 66-71. (34) Winlaw, D. S., Smythe, G. A., Keogh, A. M., Schyvens, C. G., Spratt, P. M., and Macdonald, P. S. (1995) Nitric oxide production and heart failure. Lancet 345, 390-391. (35) Pannala, A. S., Rice-Evans, C. A., Halliwell, B., and Singh, S. (1997) Inhibition of peroxynitrite-mediated tyrosine nitration by catechin polyphenols. Biochem. Biophys. Res. Commun. 232, 164168. (36) Pannala, A. S., Razaq, R., Halliwell, B., Singh, S., and Rice-Evans, C. A. (1998) Inhibition of peroxynitrite dependent tyrosine nitration by hydroxycinnamates: nitration or electron donation? Free Radical Biol. Med. 24, 594-606. (37) Leeuwenburgh, C., Hardy, M. M., Hazen, S. L., Wagner, P., Ohishi, S., Steinbrecher, U. P., and Heinecke, J. W. (1997) Reactive nitrogen intermediates promote low-density lipoprotein oxidation in human atherosclerotic intima. J. Biol. Chem. 272, 1433-1436. (38) Douki, T., Delatour, T., Paganon, F., and Cadet, J. (1996) Observation and prevention of an artefactual formation of oxidized DNA bases and nucleosides in the GC-EIMS method. Carcinogenesis 17, 347-353. (39) Ravanat, J.-L., Turesky, R. J., Gremaud, E., Trudel, L. J., and Stadler, R. H. (1995) Determination of 8-oxoguanine in DNA by gas chromatography-mass spectrometry and HPLC-electrochemical detection: overestimation of the background level of the oxidized base by the gas chromatography-mass spectrometry assay. Chem. Res. Toxicol. 8, 1039-1045. (40) Ames, B. N., Shigenaga, M. K., and Hagen, T. M. (1993) Oxidants, and antioxidants, and the degenerative diseases of aging. Proc. Natl. Acad. Sci. U.S.A. 90, 7915-7922. (41) Shigenaga, M. K., Gimeno, C. J., and Ames, B. N. (1989) Urinary 8-hydroxy-2′-deoxyguanosine as a biological marker of in vivo oxidative DNA damage. Proc. Natl. Acad. Sci. U.S.A. 86, 96979701. (42) Bible, K. C., Boerner, S. A., and Kaufmann, S. H. (1999) A onestep method for protein estimation in biological samples: nitration of tyrosine in nitric acid. Anal. Biochem. 267, 217-221.

TX0200414

Artifactual Nitration Controlled Measurement of Protein-Bound 3

Recommend Documents