(SMSR) in Digestion of Human Plasma Peptides - ACS Publications

Nov 11, 2008 - BD Diagnostics, and BD Medical, One Becton Drive, Franklin Lakes, New Jersey 07417. Received May 30, 2008. Human plasma and serum ...
0 downloads 0 Views 253KB Size
Intrinsic Peptidase Activity Causes a Sequential Multi-Step Reaction (SMSR) in Digestion of Human Plasma Peptides Jizu Yi,*,† Zhaoxia Liu,† David Craft,† Patrick O’Mullan,† Gang Ju,‡ and Craig A. Gelfand*,† BD Diagnostics, and BD Medical, One Becton Drive, Franklin Lakes, New Jersey 07417 Received May 30, 2008

Human plasma and serum samples, including protein and peptide biomarkers, are subjected to preanalytical variations and instability caused by intrinsic proteases. In this study, we directly investigated the stability of peptide biomarkers by spiking an isotopically labeled peptide into human plasma and serum samples and then monitoring its time-dependent change. Fibrinogen peptide A (FPA) was used as a model substrate, and its degradation in a conventional serum and plasma either with citrate, heparin, or EDTA as the anticoagulant, or EDTA plus protease inhibitors (inhibited plasma), was measured using time-course MALDI-TOF MS analysis. The FPA and other peptides tested in this study vary in these samples. However, the peptides are most stable in the inhibited plasma followed by, in general order, EDTA plasma, citrate plasma, heparin plasma and serum, demonstrating the benefit of plasma versus serum, and protease inhibitors for biomarker stabilization. Kinetic analysis indicates that intrinsic peptidases cause an observed first-order Sequential Multiple-Step Reaction (SMSR) in digestion of the peptide. Modeling analysis of the SMSR demonstrates that step reactions differ in their kinetic rate constants, suggesting a significant contribution of the truncated end residue on the substrate specificity of the intrinsic peptidase(s). Our observations further show that synthetic peptides introduced into plasma as internal controls can also be degraded, and thus, their (in)stability as a preanalytical variable should not be overlooked. Keywords: peptide biomarker • internal standard • FPA stability • time-course MS • serum and plasma • sequential multi-step reaction (SMSR)

Introduction Human serum and plasma contain a wealth of information relating to disease status and have received significant attention as a source of potential diagnostic markers, especially within the past decade.1,2 The discovered biomarkers contain a broad spectrum of proteins3-7 and peptides.8-10 One such peptide in blood is fibrinogen peptide A (fibrinopeptide A or FPA), the level of which in recent reports has been correlated to several diseases including urothelial cancer,11 ovarian cancer,12 gastric cancer,13 hepatocellular carcinoma,14 and myocardial infarction.15 The discovery of disease markers in blood fluid continues to accelerate as proteomics technology becomes both more powerful and more widely available.16,17 In contrast, there has been notably less success in transitioning these discoveries into clinical utility,18,19 spurring a growing interest in understanding the barriers to this transition.20-23 Among the many challenges associated with protein biomarker research, preanalytical variability is a particularly complex issue, especially with blood samples,24-26 but comparatively few studies have focused on thoroughly investigating the * To whom correspondence may be addressed at 1 Becton Drive, MC 305, Franklin Lakes, NJ 07417. Tel: (201)-847-5462 (J.Y.), (201)-847-4036 (C.A.G.). E-mail: [email protected] (J,Y,), [email protected] (C.A.G.). Fax: (201)847-4851. † BD Diagnostics. ‡ BD Medical.

5112 Journal of Proteome Research 2008, 7, 5112–5118 Published on Web 11/11/2008

biochemical processes underlying the variability. A freshly drawn blood sample is a living tissue, with a wide range of naturally active and ex vivo activated biochemical pathways, which may alter or destroy protein and peptide content in the sample. Some of the identified blood enzymes (e.g., proteases) are involved in common biochemical processes, such as the protease cascade responsible for coagulation;27 while others’ functions are not fully known. Our recent work has demonstrated extensive protease and peptidase activity in typical blood samples,20 suggesting that the inclusion of protease inhibitors during blood collection provides a more robust sample for biomarker discovery. At the same time, efforts toward establishing simple, costeffective, reproducible and accurate protein biomarker measurements also continue to dominate proteomics research activities. Currently, there is an active focus on proteomic methods that give accurate quantitation of analytes, either in relative or absolute terms. One currently used approach, “absolute quantitation” (coined AQUA),28-30 relies on spiking stable isotopically labeled peptides into samples and then mass spectrometry analysis. We adopted this methodology to investigate the stability of natural peptides that had been used, or might potentially be used, as disease markers. The peptide stability is examined in various common blood samples to assess the effect of the anticoagulants and protease inhibitors. By focusing on FPA, we find that the isotopically labeled peptide is unstable in blood samples due to 10.1021/pr800396c CCC: $40.75

 2008 American Chemical Society

research articles

Plasma Peptide Digestion by Intrinsic Peptidases intrinsic peptidase activities. We report not only the loss of the full-length AQUA FPA, but also the observed pathway of its degradation by tracking the truncated isotopically labeled peptides over time. Further kinetics and modeling analysis indicates that the degradation of peptides fits to a first-order, sequential multistep reaction (SMSR). The implications of these observations for the utility of endogenous peptides as biomarkers and spiked peptides as internal standards in quantitative methods are discussed.

Materials and Methods Blood Collection and Plasma/Serum Preparation. Human blood from healthy individuals (informed consent was obtained from all subjects) was directly drawn into evacuated tubes to obtain serum and plasma samples, as described previously.20 Briefly, serum tubes (BD product # 366430) were placed at room temperature (r.t., approximately 24 °C) to clot for 60 min after the collection of the blood, and then centrifuged at 2500g for 15 min at r.t. Plasma tubes, including citrate tube (BD product # 369714), heparin tube (BD product # 367886), EDTA tube (BD product # 367525), and BD P100 (For research use only. Not for use in diagnostics procedures.) tube (BD product # 366456), were spun immediately (within 10 min after blood was drawn), for 15 min at 2500g, and at r.t. to minimize the handling condition and to prevent possible platelet activation. Both plasma and serum samples were pipetted out of the bloodcollection tubes into Eppendorf tubes and were frozen within 15 min at -80 °C until use. With these optimized handling processes, full acquisition and processing time is minimized to approximately 30 min for plasma and 90 min for serum. AQUA Peptide. Isotopically labeled peptides were purchased from Sigma-Genosys (The Woodlands, TX). All peptides were synthesized with one stable isotopic amino acid residue, either 13 15 13 15 L-Arginine- C6 N4, or L-Phenylalanine- C9 N1, with mass of 10 Da higher than its natural counterpart. The quality of each peptide was verified by HPLC, MALDI-TOF MS (Supporting Information (SI): Figure S1), and amino acid analysis by the vendor with purity higher than 98%. Easily aqueous soluble peptides, including C4 (NGFKSHALQLNNRQIR) and FPA (ADSGEGDFLAEGGGVR) (the stable isotopic labeled residue is underlined, same hereinafter), were dissolved and diluted in water with 0.1% TFA. Peptides of lower solubility, such as C3f (SSKITHRIHWESASLLR), were dissolved in 50% ACN solution. The peptides were diluted with relevant solvents to make stocks of 400 fmol/µL for spiking experiments. Sample Preparation for Time-Course MALDI-TOF Mass Spectrometric Analysis. The frozen serum and plasma were thawed (once only) in a r.t. water bath for approximately 5 min, in order to minimize the dwell time by temperature exposure of the samples. Soluble AQUA peptide (50 µL) was spiked into a thawed serum or plasma sample (450 µL), resulting in a final concentration of 40 fmol/µL. The samples were incubated at room temperature. A 45-µL aliquot was withdrawn at each specified time, and quenched by adding 5 µL of 2% TFA solution. The quenched sample was subsequently transferred onto a Microcon YM-3 (Millipore) and spun in a Micro Centrifuge (Eppendorf Centrifuge 5417R) at 12 500 rpm and 10 °C for 45 min. The filtrate was collected, and desalted using Zip-Tip C18 (Millipore). The eluted peptides (1 µL) were mixed in 1:1 (v/v) ratio with 5 mg/ mL of R-cyano-4-hydroxycynnamic acid (CHCA) as the matrix. The peptide mixture was spotted on a plate, air-dried, and analyzed using MALDI-TOF MS.

For the less soluble peptides, 1 µL of peptide stock was added into 9 µL of serum or plasma. After incubation for a specified time period (0-72 h), the sample was quenched with addition of 40 µL of acetonitrile (ACN) with TFA solution to a final concentration of 10% ACN and 0.2% TFA. The quenched samples were supplied for extraction using Zip-Tip C18 (Millipore), followed by MALDI-TOF MS analysis. MALDI-TOF MS. MALDI-TOF MS analysis was performed on an Ultraflex II MALDI-TOF MS (Bruker-Daltonics) as described previously.20 The final spectrum was calibrated externally, which allows a mass accuracy of better than 10 ppm. For quantitative analysis of AQUA peptides, the spectrum of each sample was obtained from accumulation of 30 qualified spectra, each of which was obtained from 100 laser shots under fixed laser power. The sample site targeted by the laser was moved automatically after each of 100 shots to prevent the sample from being overburned. During the accumulation, the quality of each spectrum from the 100 shots was evaluated in terms of peak resolution and signal-to-noise ratio. A spectrum with less than 1000 peak resolution was filtered out. Under these settings, the CV% of a peptide peak intensity was 6-25% (average 15%) with this high resolution instrument,20 better than average 18% CV reported previously with a Bench-Top instrument.31 Spectral Analysis. The MALDI-TOF mass spectra were processed by flexAnalysis (Bruker-Daltonics) with median smoothing and baseline subtraction. The peaks were detected with SNP algorithm and S/N threshold 3. The other parameters were the same as those in the default method. For comparative analysis, all of the time-course spectra were processed using the same parameters and the spectra were plotted in the same scale on x- and y-axes. Kinetic Analysis. The stability of a peptide was analyzed by time-course MS-based kinetics analysis. The peak list including peak intensity was exported into Microsoft Excel using flexAnalysis (Bruker-Datonics). A roughly linear relationship is observed between the log of peak intensity versus reaction time, suggesting that degradation occurs, to a first approximation, according to a first-order reaction. The kinetic rate constant can be determined by fitting the data with: Ln(I) ) -kobst + C, where I is the peak intensity, or the peak area, t is the incubation time period in this study, C is a constant, and kobs is the observed (or apparent) rate constant. Further, the halflife of the peptide is determined by t1/2 ) Ln(2)/kobs. For determining kinetic rate constants in a complex sequential reaction observed in this study, analytical solutions were derived from our proposed model reaction, a first-order sequential multi-step reaction (SMSR), as described in SI Appendix. MATLAB (Version 7.0, Mathworks) was employed to numerically simulate this model with the experimental data, and reaction rate constants were then determined in the sequential pattern following the chain reaction sequences. The modeling process was optimized to achieve a best fit to R2 (goodness-of-fit) calculation.

Results A direct test for stability of a peptide biomarker in a collected blood sample was carried out by monitoring its change during a time-course incubation. Our original intention was to use stable isotope-labeled peptides as normalization controls, against which to compare variable and/or changing levels of intrinsic peptides. We found that the spiked peptides themselves are subject to preanalytical degradation, and thus the Journal of Proteome Research • Vol. 7, No. 12, 2008 5113

research articles spiked peptides became our focus for studying the relative (in)stability of peptides in various blood samples. We tested the stability of AQUA FPA by adding it into either a serum or plasma sample [either BD P100 (for research use only; not for use in diagnostics procedures), EDTA, citrate, or heparin plasma] collected from a single venipuncture, or a pooled serum or plasma sample from three healthy individuals. After the spiked sample was incubated for specified periods of time, the peptides were measured by MALDI-TOF MS. The change in labeled FPA (1546.69 m/z) abundance during time-dependent incubation, according to sample types (P100, EDTA, citrate, and heparin plasma samples), are shown with peak intensities normalized to ion count in Figure 1. A decrease in peak intensity of the AQUA FPA is observed in all of these plasma samples, eventually achieving undetectable levels after some period of incubation. The peptide cannot be detected above noise after 4 h in heparin, 8 h in citrate, or 24 h in EDTA. By contrast, the same peak is still easily detected at 48 h in P100, suggesting that the protease inhibitors in P100 enhance the stability of the spiked AQUA peptide. To demonstrate the quantitative nature of this analysis for monitoring protease-catalyzed reactions, we examined the dynamic range of these isotope-labeled peptide measurements. A reasonably linear plot (R2 ) 0.95) of peak intensity versus the concentration of AQUA FPA spiked into the same P100 plasma samples in four replicates was observed with its concentration up to 100 fmol/µL (SI Figure S2A). A similar linearity was also observed on tested C3f peptide (SI Figure S2B), bradykinin and GLP-1 (data not shown). This analysis supports the use of our method for semiquantitative measurement across approximately 1 order of magnitude in a complex sample such as plasma.32 As shown in Figure 2, the peak intensity (in nature logarithmic scale) of the parent labeled FPA peptide spiked into four plasma samples deceases in a linear manner as a function of reaction time at r.t. The linearity of these plots indicates that AQUA peptide degradation follows a first-order reaction of kinetics, regardless of the sample types and peptides (Table 1) tested in this study, and thus, the slopes of the plots reflect the apparent rate constants (k) of the degradation reactions. These rate constants and the corresponding half-lives of the spiked peptides are determined and listed in Table 1. Interestingly, the half-life of the FPA peptide in serum is too short (less than 15 min, Table 1) to be measured in this format. Yet, the presence of protease inhibitors in P100 provides the highest level stability to the peptides, with stability of the other samples being EDTA > Citrate ∼ Heparin > Serum. The observed decrease of the spiked AQUA FPA peptide is caused indeed by the intrinsic amino-peptidase activity,9,15,20,33 as its six fragments, sequentially truncated from the amino terminus, can be detected during the incubation (Figure 3 and SI Figure S3). The process is easily observed in heparinized plasma, since the relative instability in this sample facilitates detection of the various truncated peptides within a reasonable experimental time frame (Figure 3). At “Time 0”, which means exposure for only a few seconds of the labeled peptide to the plasma prior to quenching, the two largest truncated (“daughter”) peptides (FPA-1, and FPA-2) were detected (Figure 3), suggesting rapid enzymatic degradation. Note that the starting FPA prior to spiking is measured as a clean, single peptide with expected molecular weight (SI Figure S1), and thus, even a few seconds of exposure in the heparin plasma is sufficient to demonstrate the degradation of the spiked peptide. During 5114

Journal of Proteome Research • Vol. 7, No. 12, 2008

Yi et al. further incubation, smaller daughter peptides can be detected: FPA-3 at 30 min, FPA-4 at 1 h, and both FPA-5 and FPA-6 at 2 h (Figure 3 and SI Figure S3). None of these peptides can be detected after 8-h incubation. Importantly, the abundance of the first four daughter peptides (FPA-1 to FPA-4) displays a pattern of increased intensity up to about 2 h and decreases thereafter (Figure 3). This increase-then-decrease pattern, depicted in Figure 4B, clearly indicates that these truncated peptides are intermediates of a sequential (or consecutive) reaction caused by intrinsic plasma amino-peptidase activity. The measurement of the degradation products is sufficiently unambiguous as to allow definition of a kinetic model of the degradation process. To determine the rate constants of the sequential reaction, the fragments of FPA: FPA-1, FPA-2, FPA3, FPA-4 (Figure 3), were measured in heparin plasma pooled from three healthy individuals with four replicates, and the average of peak intensities were modeled according to a firstorder SMSR (Figure 4 and detailed in SI Appendix). Simulation of the peak intensities of time-course MS (Figure 4B) gives apparent rate constants of the sequential reactions: k1 ) 2.1, k2 ) 0.29, k3 ) 1.96, k4 ) 1.6, and k5 ) 0.04 h -1 for the first five step truncations (Figure 4A), with goodness-of-fit (R2) of 0.84, 0.78, 0.77, 0.87, and 0.76, respectively. Accordingly, the half-lives of FPA, FPA-1, FPA-2, FPA-3, and FPA-4 are 0.29, 2.1, 0.31, 0.38, and 15.2 h, respectively (Figure 4A). These data suggests that each peptide has a unique half-life, or a unique liability of the N-terminal residue (A, D, S, G, and E on FPA, FPA-1, FPA-2, FPA-3, and FPA-4, respectively) to further truncation. Two shorten peptides, FPA-1 and FPA-4, both with a negatively charged N-terminal residue, have enhanced halflives of 7- and 52-fold compared to the parental FPA, respectively, while other three neutral N-terminal peptides, FPA, FPA2, and FPA-3, have relatively similar half-lives. These results suggest that a negative charge on the leaving residue contributes to the stabilization of the peptide, or indicate a significant contribution of the N-terminal residue on sequence-specificity of the peptidase activity. Furthermore, the simulation curves match well with the experimental data (Figure 4B). This good fitting suggests that the modeling SMSR may be applicable to a general exopeptidase-caused sequential reaction.

Discussion Intrinsic Peptidase Activity. Peptidase-induced peptide degradation appears to be fairly ubiquitous, as instability can be seen for all isotopic labeled peptides tested to date in this laboratory, including complement component 4 (C4) and component 3 peptide (C3f) (Table 1), brain natriuretic peptide (BNP), glucagon-like peptide-1 (GLP-1), and bradykinin (data not shown). Both aminopeptidase and carboxypeptidase activities are constantly observed on the spiked peptides, for example, aminopeptidases truncate the N-terminal residues of FPA and C3f, and carboxypeptidases act on the C-terminal residues of C4 peptides, similar to the observation of these exopeptidase activities acting on the intrinsic peptides.9,20,33 The degradation of these spiked peptides fits the observed firstorder kinetics, and therefore, the half-life of each peptide is, to a large extent, independent of its concentration. Thus, both abundant and rare peptides, whether spiked or naturally existing in the blood samples, may be subjected to peptidasemediated instability without regard to peptide abundance. Spiking of distinctly detectable isotope-labeled peptides also gives us in vitro controls that cannot be gained otherwise. The

Plasma Peptide Digestion by Intrinsic Peptidases

research articles

Figure 1. Instability of FPA in plasma samples. Synthesized AQUA FPA peptide (for its purity, see SI Figure S1) was spiked into either P100 (A), EDTA (B), Citrate (C), or Heparin (D) plasma. The mixed samples were incubated for specified time periods as indicated on the figure. A total of 45 µL of aliquot was then removed and quenched with 5 µL of 2% TFA solution, followed by peptide extraction and MALDI-TOF MS analysis. The time-course mass spectra of the spiked FPA peptide indicate that the intensity of the peptide decreases over time-course incubation.

fast change in intrinsic peptides makes it virtually impossible to characterize their (in)stability during early sample collection.20 Spiking the uniquely identifiable exogenous control allows us to define a true “time zero” for time-dependent studies, as well as to track the spiked peptide and its breakdown

products which can be easily distinguished from endogenous peptides by MS-based analysis. Compared with previous works9,15,20,33 showing “what” happen in terms of peptidasecaused truncations of endogenous peptides, our current kinetics analysis further provides “how” the peptides get cleaved. Journal of Proteome Research • Vol. 7, No. 12, 2008 5115

research articles

Figure 2. First-order degradation of peptides in plasma samples. The peak intensities of the FPA peptide in natural logarithm are plotted versus incubation time. The experiment and date analysis were carried out in four replicates from the same individual sample of P100, EDTA, Citrate, and Heparin. One of the four replicates might be removed from analysis if this data had more than 20% deviation from the average of the other three replicates due to possible “hot-spot” of MALDI sample.

As the intrinsic peptidase(s) responsible for these truncations have been not identified, multiple peptidases and/or proteases are possibly involved in the digestion of any one peptide. In this case, the observed reaction is possibly the summation of multiple reactions, and the measured rate constant is actually an observed rate constant (kobs) of this observed reaction. Yet, our first-order SMSR model (SI Appendix) is consistent with a simple Michaelis-Menten kinetics for each step reaction: S (peptide) + E (peptidase) T SE f P (shorter peptide) + E, considering that the assumption of Km.[S] and thus, V ) Vmax[S]/Km or k ) Vmax/Km, is appropriate in this system as the concentration of each parent or daughter peptide is no more than the initial concentration of spiked peptide at 50 fmol/µL, that is 5 × 10-9 M, while a blood peptidase is expected to have much higher value of Km, such as human plasma FXIII (Km ) 1.98 × 10-5 M),34 and carboxypeptidase N (Km ) (0.25-1.5) × 10-3 M).35 Peptide Biomarker Degradation Influenced by Both Peptide Sequence and Blood Sample Type. We have observed two related features of instability with regard to the identity of the peptides, suggesting substrate specificity of the peptidase activity intrinsic to blood samples. First, while peptidasemediated instability seems to be universal, different peptides show different half-lives (Table 1), indicating an influence of peptide sequence upon intrinsic instability. Second, and more specifically, half-life differences are observed among the parent and four daughter peptides of FPA family suggesting that peptides with negatively charged amino-terminal residues have better stability (Figure 4). But the three FPA peptides, FPA, FPA2, and FPA-3, with neutral N-terminal resides A, S, and G, respectively, show only a slight, if any, difference in their halflives with 0.29, 0.31, and 0.38 h, respectively (Figure 4A). It is not clear if this difference is due to the slight differences in the peptide size/length or in the three N-terminal residues or both. However, it is reasonable to conclude that plasma and serum peptides, including potentially important and useful peptide biomarkers, vary over time in a sequence-dependent manner, and that these intrinsic peptidase(s)-induced varia5116

Journal of Proteome Research • Vol. 7, No. 12, 2008

Yi et al. tions follow the observed first-order SMSRs in digestion of peptides (Figures 2 and 4). The primary, full-length FPA (“parental” FPA) generated in serum sample has been reported to be further subjected to ex vivo peptidase digestion into shorter peptides. The pattern of the FPA family of peptides can distinguish myocardial infarction,15 and cancer diseases9 from controls, thus, implying that some disease states can alter intrinsic peptidase activity in blood, and activity measurements may have diagnostic utility. In the current study, simultaneous detection of changes among all of the FPA-derived peptides (Figure 3), and modeling according to SMSR reveals that FPA-4 has the longest half-life among this peptide family (Figure 4), at least in the heparin sample. Our observations also show that different blood sample types lead to different relative stabilities (as discussed bellow). Whether a peptide biomarker is intrinsic to blood in vivo or is generated ex vivo during the sample process, harnessing peptidase activity as the biomarker and performing a careful evaluation of stability as shown here can provide useful information for moving a newly discovered peptide successfully through biomarker validation. We also describe a strong and obvious influence of the nature of the blood sample itself (serum and various plasmas) upon the degradation process (Figure 1). The surprising speed of degradation in serum suggests that peptidase activity is stimulated during the clotting processes, further supporting the general observation of plasma being an intrinsically more stable sample than serum.20,24 Since each type of plasmas has alternative chemical (either EDTA, citrate, or heparin) included in the collection tube, the chemical must uniquely alter subsets of peptidase activities, leading to the differential stability that we observe. We expect that there are several or more different types of peptidases in blood, each being differentially inhibited by typical anticoagulants. Further inclusion of protease inhibitors in an EDTA plasma, facilitated by BD P100 with its formulation providing a broader spectrum of inhibitors against more intrinsic peptidases, provides further stabilization for plasma peptidome studies (Figure 1 and Table 1), consistent with our previous observations that the protease-inhibited plasma provides a more robust sample for proteome study.20 Furthermore, the dynamic nature of peptides and their truncated subfamilies is particularly of interest given the extent of variability in the healthy samples studied here. One implication is that researchers may benefit from studying the sampleand time-dependent stability of interested peptides in rulingin or ruling-out the candidate markers in term of their instability to enhance the likelihood of successful biomarker validation. Additionally, it also seems reasonable that use of the stabilized sample (such as the protease-inhibited plasma) that is resistant to these dynamics and variability minimizes the associated needs both for extra instability study during biomarker discovery and validation and for the even more difficult challenge of precisely controlling conditions (such as dwell time and temperature) during blood collection and handling in future routine clinical practice. Stability of Peptides Added as Internal Standards. Stable isotopically labeled peptides are often used as internal standards (or controls) for quantitative proteomics.28-30 The current study shows that the use of these control peptides should also include careful evaluation of their stability, as the instability of the spiked peptides shown here can be remarkably influenced by the nature of the blood sample itself. The fast speed at which some blood samples (e.g., serum and heparin plasma)

research articles

Plasma Peptide Digestion by Intrinsic Peptidases a

Table 1. Half-Lives of the Peptides T1/2 (h)

FPA C3f C4

serum

heparin

citrate

EDTA

P100