Profiling of Endogenous Peptides in Human Synovial Fluid by NanoLC-MS: Method Validation and Peptide Identification Jurre J. Kamphorst,†,‡ Rob van der Heijden,*,†,‡ Jeroen DeGroot,§ Floris P. J. G. Lafeber,| Theo H. Reijmers,† Benno van El,§ Ubbo R. Tjaden,† Jan van der Greef,†,‡,§ and Thomas Hankemeier†,‡ Division of Analytical Biosciences, LACDR, Leiden University, P.O. Box 9502, 2300 RA Leiden, the Netherlands, Centre for Medical Systems Biology, P.O. Box 9504, 2300 RA Leiden, the Netherlands, TNO Quality of Life, P.O. Box 2215, 2301 Leiden, the Netherlands, and Department of Rheumatology & Clinical Immunology, University Medical Center Utrecht, P.O. Box 85500, 3508 GA Utrecht, the Netherlands Received July 20, 2007
Synovial fluid potentially contains markers for early diagnosis and disease progression in degenerative joint diseases such as osteoarthritis. Here, a method is described for profiling endogenous peptides in human synovial fluid, using ultrafiltration, solid-phase extraction, nanoscale liquid chromatography, and high-resolution mass spectrometry. Synovial fluid is characterized by its high viscosity, caused by the presence of the lubricant hyaluronic acid. The method proved to be capable of eliminating the high concentrations of hyaluronic acid, which appeared to be necessary to obtain satisfactory analytical performance, that is, within-day relative standard deviations of 5-15%, between-day relative standard deviations of 6-16%, a linear response of R2 ) 0.994, a limit of detection in the femtomole range, and reproducible recoveries of 14-67%. With the developed method, in a synovial fluid sample from an osteoarthritis patient and a healthy control, in total, 501 peptides originating from 40 proteins were identified. Peptide cleavage products from six proteins that have been associated with osteoarthritis in earlier studies (collagen II, proteoglcycan 4, serum amyloid A, tubulin, vimentin, and Matrix Gla) could also be identified with our profiling method. The robustness of the method indicates that it can be applied in systems biology approaches for further studies on degenerative joint disease, eventually leading to a better understanding of the disease and its therapy, as well as the development of novel biomarkers to monitor these processes. Keywords: Peptide profiling • Peptidomics • NanoLC-MS • Biomarker • Synovial fluid • Osteoarthritis • Highresolution MS • Orbitrap
1. Introduction Osteoarthritis (OA) is a degenerative joint disease that is characterized by progressive cartilage destruction and bone changes, occasionally accompanied by synovial inflammation.1 While it affects the quality of life of millions, there are currently no disease-modifying treatments available, with the only existing therapies primarily comprising analgesics. To develop novel therapies, it is crucial to have a better understanding of the underlying pathophysiology. Furthermore, adequate biomarkers are needed for monitoring disease progression and efficacy of therapies.2 Synovial fluid is a potential compartment for such biomarkers in osteoarthritis, as it is derived directly from the site where the disease occurs. A group of compounds possibly providing good biomarkercandidates are the (endogenous) peptides, also called the * Corresponding author. E-mail,
[email protected]; phone, +31-71-5274320. † Leiden University. ‡ Centre for Medical Systems Biology. § TNO Quality of Life. | University Medical Center Utrecht.
4388
Journal of Proteome Research 2007, 6, 4388-4396
Published on Web 10/11/2007
peptidome or the low molecular mass proteome.3-5 Endogenous peptides play important roles in signaling and communication, for example, as hormones, growth factors, or cytokines. Furthermore, they are indicators of protease activity. Examples of peptide biomarkers are collagen telopeptides (e.g., CTX-II) for OA, osteocalcin for osteoporosis, and Pro-GRP for small-cell lung cancer.6 Many methods have been reported for peptide profiling in serum, plasma, and cerebrospinal fluid (CSF). For example, using µLC-MS, 340 low molecular mass proteins could be identified in serum.7 In another study, 804 serum peptides were identified, belonging to 359 proteins.8 In this work, no enzymatic digestion was used, and therefore, proteolytic activity could be described by analyzing peptide cleavage sites. In another article, the focus was mainly on the analytical performance of the LC-MS-based peptide profiling method.9 The authors achieved a median relative standard deviation in peptide intensities of 27%. An example of CSF peptide profiling is a study, in which 20 peptides were identified, originating from 12 unique proteins of which 4 were known to be 10.1021/pr0704534 CCC: $37.00
2007 American Chemical Society
research articles
Profiling of Endogenous Peptides in Human Synovial Fluid
associated with CNS disorders.10 Svensson et al. identified some 20 novel neuropeptides from mouse and rat hypothalamic extracts using ultrafiltration and nanoLC-MS.11 While targeted approaches for the detection of collagen type II neoepitope peptides by an immunoaffinity LC-MS2 assay and for serum amyloid A-derived peptides by LC-MS have been reported in literature,12,13 to our best knowledge, no method has been reported to generate comprehensive untargeted peptide profiles from synovial fluid (SF), the lubricating fluid that is found in the joints and that provides the cells in the cartilage with nutrition and regulatory signals.1 Peptides are of interest for OA research, because both cell signaling and protein degradation are known to be aberrant in OA. For a peptide profiling method to be successful, two criteria need to be fulfilled: the sample preparation method needs to be selective, removing large biomolecules that interfere with the analysis,14 and the analysis must be very sensitive, able to measure femtomoles or less. For sample preparation of peptides, multiple techniques exist.15 They can either be classified as ‘hard’, with the conditions being quite different from physiological conditions, or ‘soft’, with conditions being similar to physiological conditions. Examples of ‘hard’ techniques are liquid-liquid extraction (LLE), solid-phase extraction (SPE), and precipitation, and examples of ‘soft’ techniques are dialysis, ultrafiltration, and centrifugation. In practice, the sample preparation techniques that are used most frequently for peptide profiling studies are ultrafiltration, SPE, and precipitation. Aristoteli et al. compared these preparation techniques for plasma and found that SPE performed best with respect to the number of peptides detected and the mass range of peptides observed by MALDI.16 However, it was noted that protein ‘contaminants’ were absent only when, prior to SPE, a preparation technique as ultrafiltration or precipitation was performed. As part of our systems biology approach to OA, with which we try to understand the disease from the level of molecular pathways and the structure and dynamics of regulatory networks,17-19 the aim was to develop a validated method for the untargeted analysis of peptides in synovial fluid. We found that the analysis of peptides is seriously hampered by the presence of hyaluronic acid (HA). HA is a large carbohydrate polymer that is present in 1-3 mg/mL concentrations20 and has the function of giving SF a high viscosity, so that it can act as a lubricant. Therefore, it was investigated whether a ‘soft’ and/or a ‘hard’ sample preparation technique was needed to generated peptide profiles from SF. As SF sample volumes are often limited, especially as we want to compare clinical studies with animal models, and sensitivity is very important, nanoLC with an eluent flow rate of ∼150 nL/min was used. Theoretically, reducing a column’s internal diameter from, for example, 4.6 mm to 50 µm results in a concentration sensitivity gain factor of 8500, assuming that the same amount of sample is used.21 While in practice the increase in sensitivity is less than expected, it is still very significant, for example, when reducing the column internal diameter from 4.6 to 0.25 mm resulted in a 163 times larger signal, rather than the factor of 339 as theoretically expected.22 The developed method was validated with regard to selectivity (effectiveness of HA removal), repeatability, linearity and limit of detection, ruggedness (effect of supplementing SF with HA on peptide recovery), and recovery. The validated method
was then used for identification of peptides in SF from a patient with OA and a healthy control.
2. Materials and Methods 2.1. Chemicals. ULC/MS grade acetonitrile (ACN), formic acid (FA, 99%), and ULC/MS grade water (for nanoLC analysis) were purchased from Biosolve (Valkenswaard, The Netherlands). TFA (99%) was purchased from Sigma (St. Louis, MO). Dimethyl sulfoxide (DMSO) and ammonium bicarbonate were purchased from Baker (Deventer, The Netherlands). Ultrapure water (5.5 µS/m), used for sample preparation, was obtained from a Milli-Q gradient A10 system (Millipore, Bedford, MA). C-peptide was obtained from Genscript (Piscataway, NJ), hyaluronic acid from Fluka (Buchs, Switzerland), and the hyaluronic acid ELISA assay from Echelon (Salt Lake City, UT). 2.2. SF Samples. SF samples were obtained from the knees of patients with rheumatoid arthritis (RA, via arthrocentesis), osteoarthritis (OA, during surgery), and from an individual without a joint disease (postmortem) and were stored at -80 °C. All procedures were conducted according to local ethical standards. The SF from one RA patient was used for the method validation, as only the volume taken from this patient was sufficient for performing all experiments. The SF from one OA patient and one postmortem sample (no joint disease) were used for the peptide identification experiment. 2.3. Sample Preparation. Fifty microliters of thawed SF was diluted in 230 µL of 50 mM ammonium bicarbonate solution, pH 8, containing 5% DMSO. To this 20 µL, 10% DMSO containing 400 nmol/L of the internal standard C-peptide was added. The sample was then centrifuged for 10 min at 14 000g and 4 °C, after which the supernatant was transferred to a prewashed ultrafiltration device (Microcon YM-30, Millipore, Bedford, MA) with a nominal molecular mass limit of 30 kDa, and centrifuged at 10 000g for 50 min at 4 °C. Subsequently, 100 µL of the ammonium bicarbonate solution was added on top of the membrane to wash the residual material and to extract remaining peptides, after which the device was centrifuged again for 10 min. The resulting filtrate of about 400 µL was acidified with 50 µL of 10% TFA, and SPE was performed with POROS RP-2 C18 TopTips (Glygen, Columbia, MD) according to the manufacturer’s instructions (the peptides were eluted with 70% ACN, containing 0.1% TFA). The eluate was lyophilized, reconstituted in 20 µL of Milli-Q with 5% FA and 5% DMSO,23 and stored at -80 °C until analysis. 2.4. Reversed-Phase NanoLC-MS. For separation, an Ultimate 3000 pump system (LC Packings, Amsterdam, The Netherlands) was used, in combination with an LTQ XLOrbitrap mass spectrometer for the identification experiment and an LTQ-FT mass spectrometer for the other experiments (Thermo Scientific, San Jose, CA). To have an efficient nanoLCMS system, the system’s dead volume, both between the precolumn and column (precolumn dead volume) and after the column (postcolumn dead volume) was minimized. This was achieved by using a Dean’s switching setup as described by Meiring et al.21 In our particular case, 4 µL of sample was loaded onto a precolumn (0.1 × 20 mm, 5 µm C18 particles; NanoSeparations, Nieuwkoop, The Netherlands) with a flow rate of 5 µL/min using 0.1% FA. After loading and washing for 5 min, the precolumn and column (0.05 × 270 mm, 3 µm BioSphere C18 particles, NanoSeparations) were switched in Journal of Proteome Research • Vol. 6, No. 11, 2007 4389
research articles series, and the peptides were eluted with a flow rate of about 150 nL/min into the mass spectrometer via a small gold and carbon sputtered spray-tip (NanoSeparations), to which the voltage is applied. In this setup, the outlet of the column and the spray-tip are directly connected to each other, resulting in a postcolumn dead volume of as little as 5 nL. The elution was performed with a linear gradient of 30 min, running from 100% solvent A (0.1% FA in H2O) and 0% solvent B (0.1% FA in ACN), to 55% (v/v) solvent A and 45% solvent B. Before the next analysis, both the precolumn and the column were first washed with 90% solvent B for 5 min and then equilibrated with 100% solvent A for 10 min. ESI and LTQ mass spectrometer settings were as follows: T, 180 °C; capillary voltage, 32 V; tube lens, 135 V; spray voltage, 2.4 kV; multipole 00 offset, -3.5 V; intermultipole lens 0 voltage, -6 V; multipole 0 offset, -5 V; intermultipole lens 1 voltage, -9 V; gate lens voltage, -60 V; multipole 1 offset, -10 V; multipole RF amplitude, 400 V p-p; front lens, -5.5 V; maximum injection time, 200 ms; number of µscans, 2; scan range, 400-1500 m/z; data were recorded in centroid mode. 2.5. Peptide Identification. For the identification of peptides, an LTQ XL-Orbitrap mass spectrometer was used. The Orbitrap settings were as follows: FT main RF amplitude, 1450 V p-p; transfer RF amplitude, 500 V p-p; storage multipole RF amplitude, 334 V p-p; ion energy, 1112 V; lens 1, 489 V; lens 2, 0 V; lens 3, -200 V; lens 4, 0 V; deflector, 294 V; general electrode, -3450 V; maximum injection time, 500 ms; number of µscans, 1. During analysis, the mass spectrometer continuously performed scan cycles in which first a high-resolution (100 000) full scan (400-1500 m/z) in profile mode was made by the Orbitrap, after which MS2 spectra were recorded in centroid mode for the 3 most intense ions (isolation width, 4 m/z; normalized collision energy, 35% and 40%). Dynamic exclusion was enabled (repeat count, 1; repeat duration, 30 s; exclusion list size, 500; exclusion duration, 180 s; relative exclusion mass width, 5 ppm) as was charge state screening. Singly charged ions were not fragmented. In addition to the 30 min gradient as discussed in the previous section, the samples were separated with a 3 h gradient running from 100% solvent A and 0% solvent B, to 72% (v/v) solvent A and 28% solvent B. The obtained spectra were correlated with a recent version of the human FASTA database (April 2007) using Bioworks v3.3 (Thermo Scientific, San Jose, CA). The following post-translational modifications (PTMs) were taken into account: C-terminal amidation; phosphorylation on S, T, and Y; hydroxylation on P and K; oxidation on M; pyroglutamic acid formed from Q or E. The following settings were used for the search engine: mass type, monoisotopic precursor, and fragments; enzyme, no enzyme; peptide tolerance, 20 ppm; fragment ion tolerance, 1 AMU; number of results scored, 250; ions and ion series calculated, B and Y; peptide matches reported, 10; PTMs per peptide, 3. For a peptide to be considered as identified, it had to match the following criteria: ∆ppm 0.1; Peptide probability
