Development of a Highly Sensitive, High-Throughput, Mass

Company, and Monarch Lifesciences, Indianapolis, Indiana. Received May 16, 2007. Type-I procollagen aminoterminal propeptide (PINP) is a useful biomar...
0 downloads 0 Views 301KB Size
Download PDF
Development of a Highly Sensitive, High-Throughput, Mass Spectrometry-Based Assay for Rat Procollagen Type-I N-Terminal Propeptide (PINP) To Measure Bone Formation Activity Bomie Han,*,† Marci Copeland,‡ Andrew G. Geiser,§ Laura V. Hale,§ Anita Harvey,§ Yanfei L. Ma,§ Connie S. Powers,| Masahiko Sato,§ Jinsam You,‡ and John E. Hale† Integrative Biology, Lilly Research Laboratories, Eli Lilly and Company, Indianapolis, Indiana, Pathology, Lilly Research Laboratories, Eli Lilly and Company, Musculoskeletal Diseases, Research Laboratories, Eli Lilly and Company, and Monarch Lifesciences, Indianapolis, Indiana Received May 16, 2007

Type-I procollagen aminoterminal propeptide (PINP) is a useful biomarker for bone formation activity that is used to monitor treatment of bone disorders including osteoporosis. Studies with human patients under long-term therapy for osteoporosis by daily injection of parathyroid hormone (PTH) demonstrated that the circulating level of PINP at 3 months of treatment, measured by radioimmunoassay, was a good predictor for bone mineral density (BMD) at 18 months. It is important to have PINP assays for other species to elucidate processes of bone formation and for the development of new therapeutic options that can enhance bone formation activity. Currently, only a human PINP radioimmunoassay is commercially available for clinical use, which may not be cross reactive with PINP from other species. For example, rat PINP has little amino acid sequence homology to human PINP. Therefore, we developed a new, highly sensitive, high-throughput mass spectrometry-based assay for PINP from rat plasma or serum that does not rely on antibody reagents. Circulating levels of PINP showed age-dependent changes in rats. Prednisolone treatment, which is known to retard bone formation activity, led to a significant decrease in PINP, whereas PTH treatment dose-dependently increased PINP. The throughput of the assay parallels that of most antibody-based assays so that it can handle a large number of samples that are generated from preclinical animal studies. PINP in rats may serve as a biomarker for bone formation activity, and this assay could be instrumental in studying bone physiology in rat experimental models. Keywords: biomarker • mass spectrometry • multiple reaction monitoring (MRM) • ion trap • proteomics • osteoporosis • PTH • PINP

Introduction Bone degenerative diseases such as osteoporosis occur in a substantial portion of the senior adult population. Osteoporosis encompasses a heterogeneous group of disorders that represent a major risk for bone fractures, and is a growing, substantial burden on the health care system. Billions of dollars are spent annually on medical care for the treatment of osteoporosis. Clinically, osteoporosis is characterized by diminished bone mass (reduced BMD) and loss of bone architecture, resulting in decreased bone strength and increased risk of bone fracture.1 Rat models of postmenopausal osteoporosis have been utilized for pharmacological investigations and have been * To whom correspondence should be addressed. Bomie Han, Integrative Biology, Eli Lilly and Company, Lilly Research Laboratories, Lilly Corporate Center, Indianapolis, IN 46285; E-mail, [email protected]; Tel, 317-4335189; Fax, 317-655-6236. † Integrative Biology. ‡ Monarch Lifesciences. § Musculoskeletal Diseases. | Pathology.

4218

Journal of Proteome Research 2007, 6, 4218-4229

Published on Web 10/09/2007

shown to be predictive of clinical efficacy for developing therapies.2 In fact, regulatory agencies require rat data as part of the approval process for a new drug application. As an example, human parathyroid hormone (PTH) 1-34, a boneformation therapy for the treatment of osteoporosis, was extensively analyzed in rat models before approval as therapy. PTH has been shown to increase bone mass, improve spatial architecture, and significantly reduce the risk of vertebral and nonvertebral fractures by 65 and 53%, respectively, in postmenopausal women with osteoporosis.3 Serum markers for bone formation activity include bone specific alkaline phosphatase, osteocalcin, PINP (procollagen type-I N-terminal propeptide), and PICP (procollagen type-I C-terminal propeptide). Bone-specific alkaline phosphatase is a plasma membrane enzyme produced by preosteoblasts and osteoblasts and serves as an early marker of osteoblast proliferation and activity.4 However, the bone alkaline phosphatase gene product is produced also in liver and kidney making specific measurement of bone-derived alkaline phosphatase difficult. An organ-specific immunoassay exists for human bone 10.1021/pr070288s CCC: $37.00

 2007 American Chemical Society

Rat PINP as a Biomarker for Bone Formation

alkaline phosphatase that is likely attributable to specific posttranslational modifications,5 but similar assays for other species do not exist. Osteocalcin is a protein synthesized by osteoblasts and deposited into the mineralized bone matrix and serves as a late phase marker for bone formation activity.6 However, circulating levels of osteocalcin may also increase upon stimulation of osteoclastic bone resorption activity due to release of osteocalcin from the bone matrix.7 Both PINP and PICP are produced from type-I procollagen during proteolytic processing of the procollagen into tropocollagen fibrils. Because type-I collagen comprises 90% of the protein in bone matrix, circulating levels of PINP and PICP may serve as a direct indicator for bone matrix formation. Despite the common origin of PINP and PICP, PINP was shown to be a superior indicator of bone formation activity,8 especially for PTH activity.9 Human studies with PTH indicate that circulating levels of PINP predict long-term clinical efficacy in terms of vertebral BMD.9 Human PINP can be measured using a commercially available RIA kit.10 Assay methods for other species, however, are currently not commercially available. It is critical to be able to measure PINP in other species for the development of additional treatment options. Rats are common laboratory animals that are useful models for drug development.2 Therefore, we describe a highly sensitive, high-throughput, mass spectrometry-based method to measure rat PINP that does not require an antibody.

Experimental Section Reagents. Trypsin-gold was purchased from Promega (Cat # V5280). Animal serum or plasma was obtained from the following sources; dog serum, pig serum, sheep serum, goat serum, and rabbit serum from BioMeda (Cat #s BMES1034, BMES1012, BMES1017, BMES1028, BMES1010, respectively), horse serum from HyClone (Cat # SH30074.02), rat plasma from Harlan (Cat # 4511), fetal bovine serum from GibcoBRL (Cat # 10099-141), and guinea pig plasma from Lampire (Cat # P1090N-10). Sprague-Dawley rats from Harlan were kept in house before collecting serum at different ages. Rat sera from 2 year old animals were a gift from Dr. Stuart Warden at Indiana University Medical School. Ammonium carbonate, ammonium bicarbonate, 2-iodoethanol, and triethylphosphine were from Sigma. Mass-spectrometry grade formic acid was from Sigma. Water with 0.1% formic acid was from Fisher Scientific. Acetonitrile was from Burdick & Jackson. Synthetic peptides were from Midwest Biotech (Fishers, IN). Sample Preparation Prior to LC-MS/MS. Serum or plasma proteins were digested with trypsin before analysis by tandem mass spectrometry coupled in-line with high performance liquid chromatography (LC-MS/MS). Because the target peptide from the PINP contained a Cys residue, serum/plasma proteins were reduced and alkylated prior to trypsin digestion. Reduction and alkylation of the serum or plasma proteins was done in one step essentially as described earlier11 with the following modifications. Most importantly, urea was omitted during the coupled reduction/alkylation step. Typically, 10 µL of serum or plasma sample was diluted with 50 µL of ammonium carbonate buffer (0.1 M, pH 11) in a polypropylene container and kept on ice followed by mixing with 80 µL of reduction/alkylation cocktail (R/A cocktail) at room temperature. The R/A cocktail was prepared by mixing 0.5 mL 2-iodoethanol, 0.125 mL triethylphosphine, and 24.375 mL of acetonitrile (2-Iodoethanol comes with copper granules as a stabilizer and was filtered through 0.45 µm spin filter (Millipore

research articles UFC30HV00) immediately prior to preparation of the R/A cocktail). For smaller volume of samples, total volume was maintained the same by prediluting the serum with phosphate buffered saline (PBS). For larger volume of samples, each reagent volume was increased accordingly. After adding the R/A cocktail to the diluted sample in alkaline pH, the samples were mixed thoroughly and incubated for 1 h at 37 °C with constant shaking. Reduced and alkylated samples were centrifuged at 4000 rpm for 4 min then filtered through SolvInert filter plates (Millipore, MSRLN0450) to remove precipitated proteins. Solvents as well as the remaining reduction/alkylation reagents were removed from the filtrate by SpeedVac (miVac DUO concentrator from GeneVac Cat # DUC-12060-C00) typically under high heat (75 °C) for 6 h followed by an additional 1218 h at room temperature. Dried samples were dissolved in 100 µL of 100 mM ammonium bicarbonate solution (ABC) containing trypsin (1 µg of Trypsin-gold per 10 µL initial plasma or serum volume). The best results were obtained when samples were reconstituted with Trypsin-gold immediately after removal from the speedvac. Plates were sealed using pierceable heat-sealing aluminum foil (ABgene Cat # AB-0757) using a heat sealer (Eppendorf, Cat # 5390) and incubated with trypsin for 6 h to overnight then filtered through SolvInert filter plates (Millipore, MSRLN0450) before injecting 50 µL to LC-MS/MS system. Optimization of the Sample Preparation Procedure for High-Throughput Handling. Reduction/alkylation reaction was performed in 96-well PCR plates with a tall raised-rim around individual wells (Robbins, Surrey UK, Cat # 1055-00-0). A precursor of an internal standard peptide (iSTDext; [pE]EDIPEVSCIHNG[L+7]RVPNGETWK, where [pE] indicates pyroglutamyl residue and [L+7] indicates Leu residue with the molecular weight 7 amu higher than the natural counterpart. See Results for details) was prepared in ice-cold ammonium carbonate buffer at 50 nM concentration. Fifty microliter of this solution was dispensed into the PCR plates using a MultiDrop (Thermo). The PCR plates were kept chilled on ice while 10 µL of serum or plasma samples were transferred and mixed in duplicate. The R/A cocktail was added at room temperature using an eight-channel multidispense pipet. Prerinsing of the pipet tips was important for accurate delivery of the reagent due to high vapor pressure of the acetonitrile in the solution. Plates were sealed using pierceable heat-sealing aluminum foil then mixed thoroughly. A large precipitation of proteins was readily visible at this step. Plates were incubated at 37 °C for 1 h with moderate shaking. Plates were centrifuged for 4 min at 4000 rpm before peeling the sealing foil. The filtration assembly was prepared by putting a SolvInert filter plate on top of the tall raised-rim PCR plate (TempPlate II from USA Scientific, Cat # 1402-9600) as a receiving plate in a locking position. The outlet of this filter plate fits into the raised rim of the receiving plate. The filtration assembly was placed over the sample plate in an upside-down position to form a filtration sandwich so that the raised rim of the sample plate is inserted into individual well of the filter plate. The filtration sandwich was inverted and centrifuged for 1 min at 1000 rpm followed by 4 min at 4000 rpm. Because the protein precipitation had formed a tight pellet during the previous spin, the filters were not clogged. The filtrates were dried by SpeedVac as described above and then samples were reconstituted with Trypsin-gold, the plates sealed and samples digested at 37 °C overnight. Because the sample preparation method involves Journal of Proteome Research • Vol. 6, No. 11, 2007 4219

research articles two filtration steps, the final sample plate is in the same orientation as the initial reduction/alkylation plate. LC-MS/MS of PINP-Derived Tryptic Peptide. N-terminal tryptic peptide derived from the alpha 1 chain of the PINP (pNTTP; [pE]EDIPEVSC*IHNGLR where C* is Cys residue with the sulfhydryl group modified with EtOH, described in detail in Results section) was measured by in-line LC-MS/MS for quantitation of PINP. In the internal standard (iSTD) peptide, the Leu residue of the pNTTP peptide is uniformly labeled with N15 and C13. Rat serum contained an unidentified peptide that had similar parent m/z and distinct but overlapping MS-MS pattern as the pNTTP peptide. This and other interfering peptides were separated by an HPLC system (Surveyor MS pump from Thermo Finnigan) on a C18 reversed-phase column (XBridge 2.5 µm × 2.1 mm × 50 mm from Waters) using the following two-solvent gradient system (solvent A, 0.1% formic acid/H2O; solvent B, 0.1% formic acid/acetonitrile) in a total run time of 3.5 min per sample: 5-21% B at 200 µL/min for 0.5 min, 21% B at 500 µL/min for 0.05 min, 21% B at 500 µL/ min for another 2 min, 21-80% B at 600 µL/min for 0.2 min, 80% B at 600 µL/min for another 0.2 min, 80-0% B at 600 µL/ min for 0.2 min, 100% A at 600 µL/min for 0.35 min, followed by 100% A at 200 µL/min for 0.05 min. The HPLC column was maintained at 50 °C, and the solvents were kept at room temperature and the samples were kept at 4 °C. Typically, 50 µL of the sample out of total volume of 100 µL was injected using a sample injection loop of 100 µL and pNTTP peptide was eluted at about 2.4 min. Two water blank samples were injected before the actual samples so that the HPLC column could reach a steady state. Typical carry-over of pNTTP peptide from previous run was less than 0.1%. Total LC-MS/MS time for one sample was 4.5 min including 1 min of sample injection time between samples. Thus, 320 samples can be processed through the LC-MS/MS in 1 day. Positive ion mass spectrometry was obtained using an LTQ ion trap quadrupole mass spectrometer equipped with an ESI source (Thermo Finnigan). The entire effluent of the column was directed to the ESI source between 2 and 3 min of HPLC run, whereas the rest was diverted away from the mass spectrometer. The instrument was tuned for the pNTTP peptide at a flow rate of 500 µL/min with 21% B by infusing the peptide using a T-split. To accommodate high flow rate, certain parameters for the instrument had to be adjusted manually including transfer capillary temperature (312 °C) and nitrogen sheath flow. All microscans were set to one microscan of 50 ms collection of ions for the trap. In the instrument method, the following parameters were used for MS-MS conditions; normalized collision energy, 21; activation Q, 0.180; activation time, 50 ms. Three MS-MS transitions were measured for both pNTTP and iSTD peptides. The m/z values of each of these transitions for pNTTP peptide are from m/z of 869.60 with isolation width of 3.4 to m/z of 635.50 (y′′11 + 2 ion; pNTTP-A transition), to m/z of 692.50 (y′′12 + 2 ion; pNTTP-B transition), and to m/z of 945.00 (y′′8 + 1 ion; pNTTP-C transition). Isolation width for each MS-MS transition was 5.00, 5.00, and 6.00, respectively. The m/z values of the corresponding MSMS transitions for iSTD peptide were 873.10 for the parent ion with isolation width of 3.4, 639.00 for y′′11 + 2 ion (iSTD-A transition), 696.00 for y′′12 + 2 ion (iSTD-B transition), and 952.00 for y′′8 + 1 ion (iSTD-C transition). The isolation width for each MS-MS transition was 5.00, 5.00, and 6.00, respectively. The m/z values and isolation width for the parent and each of the three were selected so that deamidation of the Asn 4220

Journal of Proteome Research • Vol. 6, No. 11, 2007

Han et al.

residue (see Results for details) in the pNTTP peptide or the iSTD peptide could be captured. Peak Integration and Curve Fitting. Peak integration was done using a processing method within XCaliber software using the following parameters: peak integration method, ICIS; smoothing points, 5; baseline window, 15; area noise factor, 1; peak noise factor, 3 for iSTD and 5 for pNTTP; constrain peak width, 5% peak height and 3% tailing factor; advanced option, repetitive noise method. Integration range for each MS-MS transition is indicated within parentheses: pNTTP-A (634.82636.82), pNTTP-B (691.36-693.36), pNTTP-C (943.96-946.96), iSTD-A (638.33-640.33), iSTD-B (694.87-696.87), and iSTD-C (950.98-953.98). Isotopic distribution and relative intensities among three transitions for each peptide was examined and was confirmed to match with those of synthetic peptides. The ratio between the pNTTP peptide and the iSTD peptide was calculated for each transition then numeric average of the three ratios was obtained as follows: normalized peak intensity (NPI) ) (PIpNTTP-A/PIiSTD-A + PIpNTTP-B/PIiSTD-B + PIpNTTP-C/PIiSTD-C)/ 3 where PI is peak intensity. NPI values for the calibration standard samples were fitted to a sigmoidal curve (NPI ) Bottom + (Top-Bottom)/(1 + 10^((logEC50-X)*(Hill Slope))) where X is the logarithm of concentration; Bottom, Top, EC50, and Hill Slope are parameters to be determined by the curve fitting of the data) using a nonlinear curve fitting function of the GraphPad Prism (GraphPad Software, Inc., San Diego, CA) with 1/Y^2 as a weighting factor. It was important to use the weighting factor to obtain calibration curve that works over the entire concentration range equally well. Animal Experiments. Intact female, 27 week-old SpragueDawley rats (Harlan) were maintained on a 12 h light/dark cycle at 22 °C with ad lib access to food (TD 2014 with 0.72% Ca and 0.61% P, 990 IU/g D3, Teklad, Madison, WI) and water. They were divided into groups of 6 rats in each group for single or combination treatments as indicated below. Prednisolone 20 mg/kg/day (1 mL/kg dosing volume) was given daily orally via gavage. Human recombinant PTH (1-38) (Cambridge Research Biochemicals, Northwich, Cheshire, UK) was dissolved in Vehicle solution (0.9% NaCl containing 2% (v/v) heat-inactivated rat serum and 0.001 N HCl) and given daily by subcutaneous injection. Serum was collected after 10 days treatment. Animals were maintained and treated in accordance with Eli Lilly and Company’s Institutional Animal Care and Use Committee policies. Rat serum osteocalcin was measured by RIA using a kit from Biomedical Technologies, Incorporated (Stoughton, MA), with modifications to a 96-well format as described previously.12 Measurement of rat alkaline phosphatase was performed as described previously in a semi-automated fashion using Hitachi 917 (Roche Diagnostics, Indianapolis, IN)13 and with modifications as recently described.14

Results Amino Acid Sequence Analysis of Rat PINP. Intact PINP protein is a heterotrimeric helix with a predicted molecular weight of 33 kD (Figure 1), which is difficult to detect by mass spectrometry with high sensitivity. Moreover, potential modifications including hydroxylation of proline residues make it difficult to capture the intact molecule by mass spectrometry. Tryptic digestion of proteins has been used extensively to render larger proteins to smaller peptides that are amenable to high sensitivity detection by mass spectrometry. Visual inspection of the amino acid sequence of the rat PINP protein indicated that the N-terminal tryptic peptide of the alpha-1

Rat PINP as a Biomarker for Bone Formation

research articles

Figure 1. Amino acid sequence of the rat PINP. Amino acid sequences are shown for N-terminal portion of the rat procollagen typeI. Type-I collagen is a heterotrimeric helix comprised of two alpha1 and one alpha2 chains. Residues indicated by the bar above the letter constitute PINP. Residues shown in inverse color in the alpha1 chain belong to the tryptic peptide that was used to develop the mass spectrometry-based assay for rat PINP, which are identical to those of mouse but different from human.

Figure 2. Glutamine vs pyroglutamate at the N-terminus of a peptide. N-terminal glutamine residue can undergo cyclization to form pyroglutamate either spontaneously or enzymatically.

chain of the PINP (NTTP1; QEDIPEVSCIHNGLR) was a good candidate for unambiguous identification of the protein and would serve as a surrogate measure of PINP levels (2 mol of the NTTP1 per mole of PINP). Other tryptic peptides were either too short, too long, or contained potential modification sites. This peptide was determined to be specific for PINP by BLAST analysis of the non-redundant protein database. PINP contains multiple Cys residues which can form both inter- and intramolecular disulfide bonds. For efficient trypsin digestion and also for unambiguous identification of tryptic peptides containing the Cys residue, it was necessary to reduce and alkylate the sulfhydryl group of the Cys residues. We used 2-iodoethanol as an alkylating agent (see Experimental Section for detail) which replaces the cysteine -SH group with a -SCH2CH2OH group. Therefore, the synthetic peptide corresponding to the N-terminal tryptic peptide contained the appropriate modification at the Cys residue (NTTP peptide; QEDIPEVSC*IHNGLR in which C* indicates the Cys residue with the sulfhydryl group modified with EtOH). Proteins and many biological peptides with predicted Nterminal Glu or Gln residue frequently undergo cyclization of these residues to form pyroglutamate (pyroGlu). This may occur either enzymatically, spontaneously, or both (Figure 2).15,16 To test if a similar conversion had occurred in PINP, we prepared another synthetic peptide in which the N-terminal Gln of the NTTP peptide was substituted with a pyroglutamate residue (pNTTP peptide; [pE]EDIPEVSC*IHNGLR in which [pE] is pyroGlu). Using both the NTTP and pNTTP synthetic peptides, we established HPLC separation conditions and tandem MSMS fragmentation patterns (Figure 3). All major MS-MS transitions were from y-ions, and therefore, had identical m/z values between the two parent peptides. The two peptides, however, were clearly separated by HPLC and had distinct molecular weights (1754 for NTTP vs 1737 for pNTTP) making identification of each peptide from tandem LC-MS/MS unambiguous even though the MS-MS fragmentation patterns of the two peptides were very similar.

Rat plasma proteins were reduced and alkylated then digested with trypsin (see Experimental Section for details). The resulting peptide mixture was subject to LC-MS/MS analysis for the presence of NTTP and pNTTP peptides. The plasma sample contained no detectable NTTP peptide (Figure 4A). Instead, a robust signal for pNTTP peptide was easily identified (Figure 4B), indicating that the N-terminal residue of the endogenous PINP has a pyroglutamyl group. Closer analysis of the MS-MS spectrum of the pNTTP peptide in the plasma revealed that m/z values for some of the y ions were higher than the synthetic peptide by one-half or one m/z unit (Figure 4B). MS-MS fragments shifted by one m/z unit were all singly charged, whereas all ions shifted by one-half m/z unit were doubly charged (Figure 4C). This fragmentation pattern was consistent with deamidation of the Asn residue preceding the Gly residue. Such deamidation of Asn is known to occur readily especially when Asn is followed by small residues such as Gly.17 To confirm that the shift in the m/z values for the MS-MS transitions that were observed from the plasma sample was indeed due to deamidation of the peptide, we prepared another synthetic peptide (pNTTP-ND; [pE]EDIPEVSC*IHDGLR) in which the Asn of the pNTTP peptide was replaced with Asp. The two synthetic peptides, pNTTP and pNTTP-ND, had the same intrinsic signal for all MS-MS transitions that were monitored (data not shown), and had the same retention time on HPLC in this initial condition although they could be resolved under optimized condition later. This synthetic pNTTPND peptide gave an MS-MS transition pattern identical to the pNTTP peptide from the rat plasma (data not shown), confirming deamidation of the Asn. It is unclear if the N-terminal conversion of Gln to pyroGlu and the deamidation of Asn residue result from sample preparation or if the circulating PINP has this form. Even when the synthetic NTTP peptide (with N-terminal Gln and internal Asn) was subjected to the same sample preparation method, the N-terminus was quantitatively converted to pyroGlu and the Asn was deamidated. The same modifications in the N-terminal tryptic fragment of PINP were found in culture supernatants from rat-derived UMR-106 and ROS-17/2.8 cells (data not shown). These cells are known to produce type-I collagen in culture. Design of Standard Peptides. Because purified rat PINP was not available to use as a calibration standard, we used a synthetic peptide to establish calibration standards for quantitation of PINP. This peptide contained eight additional amino acid residues at the C-terminus of the pNTTP peptide which Journal of Proteome Research • Vol. 6, No. 11, 2007 4221

research articles

Han et al.

Figure 3. LC-MS/MS of N-terminal tryptic peptides of PINP. Two synthetic peptides (NTTP and pNTTP) corresponding to two possible variations of the N-terminal tryptic peptide were used to establish LC-MS/MS pattern of each peptide (Top, NTTP; Bottom, pNTTP). NTTP has the native sequence of the N-terminal tryptic peptide of the alpha-1 chain of PINP whereas pNTTP has pyroglutamate in place of the glutamine at the N-terminus. Amino acid sequences of each peptide are given above each Figure. C* in both peptides indicates Cys residue with its sulfhydryl group modified with EtOH to reflect reduction/alkylation of PINP during sample preparation. (Top left) Total ion chromatogram of NTTP for MS-MS transition of m/z ) 878 to m/z ) 635. (Top right) Full MS-MS spectrum of NTTP for parent ion of m/z ) 878. (Bottom left) Total ion chromatogram of pNTTP for MS-MS transition of m/z ) 869 to m/z ) 635. (Bottom right) Full MS-MS spectrum of pNTTP for parent ion of m/z ) 869. Major ion species from MS-MS fragmentation are indicated. See Figure 4C for definition of these ion species. HPLC employed the same two-solvent system as described in Experimental Section but used a longer gradient (100% A at 100 µL/min for 1.4 min, 0-10% B at 200 µL/min for 0.1 min, 10-30% B at 200 µL/min for 1.2 min, 30-50% B at 200 µL/min for 0.7 min, 50-80% B at 200 µL/min for 0.1 min, 80% B at 200 µL/min for another 0.45 min, 80-0% B at 200 µL/min for 0.55 min, 100% A at 200 µL/min for 1.4 min, followed by 100% A at 100 µL/min for 0.1 min)

correspond to the natural amino acid sequence of the rat PINP (pNTTPext; [pE]EDIPEVSCIHNGLRVPNGETWK). Unlike the pNTTP peptide, the single Cys residue was kept unmodified in the pNTTPext peptide. Reduction and alkylation followed by tryptic cleavage of the pNTTPext would generate the same PINP N-terminal tryptic peptide as endogenous. We used a serial dilution of the pNTTPext peptide as a set of calibration standards in place of purified PINP protein. Because the pNTTPext peptide has to undergo the same reduction/alkylation and tryptic cleavage as the plasma samples to be detected, any assay-to-assay variation in the efficiency of sample preparation can be accounted for by using this peptide as a calibration standard. In addition, we also included an internal standard (iSTD) peptide in the assay design for normalization of sample-to-sample variations within an assay during the course of sample preparation and handling. The iSTD peptide is identical to the pNTTP peptide except that its single Leu residue is uniformly labeled with stable isotopes, N15 and C13, so that the iSTD peptide behaves chemically identical to the pNTTP peptide but can be distinguished by the mass spectrometer due to its increased mass. At the beginning of the assay, all samples including the calibration standards are mixed with a fixed volume of a solution containing precursor for the 4222

Journal of Proteome Research • Vol. 6, No. 11, 2007

iSTD peptide (iSTDext; [pE]EDIPEVSCIHNG[L+7]RVPNGETWK, where [L+7] indicates Leu residue with a molecular weight seven mass units higher than the natural counterpart). Reduction/alkylation and tryptic cleavage of the iSTDext peptide would generate iSTD peptide in a manner identical to generation of pNTTP from pNTTPext in the calibration standards or from PINP in the serum or plasma samples. For quantitation, the ratio of the peak area of the MS-MS transition of the unlabeled pNTTP peptide to that of the labeled iSTD obtained from the same sample was calculated. Because multiple MSMS transitions were monitored, the average value of the ratios from each transition was used. This average ratio was compared to those of the calibration samples to obtain absolute quantitation of the PINP peptide in the sample (see Experimental Section for detail). Assay Optimization for Increased Sensitivity and Quantitative Measurement of PINP. Whereas PINP could be easily detected from rat plasma obtained from a commercial source (Harlan, Cat # 4511), PINP in the plasma or serum from inhouse rats was barely measurable (data not shown). We attributed this to the difference in the age of the animals because in-house rats were 8-9 months old, whereas rat serum obtained from most vendors is from rats less than 3 months

Rat PINP as a Biomarker for Bone Formation

research articles

Figure 4. N-terminal tryptic peptide of PINP in rat plasma has N-terminal pyroglutamate and its Asn residue is deamidated. Comparison of the LC-MS-MS patterns of NTTP and pNTTP (see Results and Figure 3 for explanation) to those obtained from rat plasma indicates that only pNTTP is found in the rat plasma under the experimental condition (A and B). Left-hand panels of (A) and (B) show total ion chromatogram of three different transitions that were monitored for each peptide as indicated on the m/z values in the right-hand panels. Right-hand panels show average intensity of each MS-MS fragmentation within time window as indicated by two opposing arrows on the left-hand panels. More careful comparison of the MS-MS patterns of pNTTP with those of rat plasma reveals that m/z values of these transitions in the rat plasma are all shifted higher by one-half mass unit if they are doubly charged or by one mass unit if they are singly charged (Panels B and C). Such pattern is consistent with deamidation of the Asn residue. HPLC condition is the same as described for Figure 3. Journal of Proteome Research • Vol. 6, No. 11, 2007 4223

research articles

Han et al.

of age. In fact, serum from 3.5 week-old rats showed high concentrations of PINP (see below for complete data). Because most of our in-house rats for bone-related animal experiments are 8-9 months old by the time of blood collection, we needed to increase the sensitivity of the assay to reliably detect the low level of PINP in these animals. Because antibodies to rat PINP are not commercially available, we could not use an immunological method to enrich PINP and thus increase the sensitivity. Therefore, we explored nonimmunological methods for enrichment of PINP. The initial conditions that we employed for sample processing were to mix 10 µL of serum or plasma with 40 µL of iSTDext solution in PBS. Ten microliters of this mixture was combined with 90 µL of ammonium carbonate buffer (pH 11) to which was added 100 µL of reduction/alkylation cocktail prepared in acetonitrile. In this condition, 2 µL of plasma sample was processed through the reduction/alkylation and trypsinization and then half of the final sample was analyzed by the LC-MS/ MS method. Simply increasing the sample volume with a corresponding increase in other reagents did not increase the signal adequately because the background also increased. Some proteins precipitated during the reduction/alkylation step and were removed by filtration. We explored different concentrations of acetonitrile during reduction/alkylation step to remove plasma proteins by precipitation while keeping PINP in solution. Increasing the acetonitrile concentration up to 62.5% gave a modest increase in the signal intensity (Figure 5A), which was attributed to removal of more proteins by precipitation without loss of PINP. Consistent with this interpretation, a significant decrease in the background noise was also observed (data not shown). Further increases in the acetonitrile concentration gave a significant decrease in the recovery of the endogenous PINP without affecting recovery of the internal standard peptide (Figure 5A). Therefore, we defined this condition (2 µL plasma in total of 60 µL aqueous sample, mixed with 100 µL of reduction/alkylation reagent in acetonitrile) as yielding full recovery of PINP and used this condition for comparison to optimize assay sensitivity (first groups in Figure 5B and C). Simple scale-up of this condition to accommodate a larger sample volume was impractical because this would increase total volume in excess of the capacity of 96-well plates and make handling large numbers of samples extremely difficult. We explored a wide range of conditions for different volumes of plasma, ammonium carbonate, and acetonitrile and also for trypsin concentrations to optimize PINP recovery and background. Increasing the plasma volume to 10 µL under the same condition increased the signal moderately (second group in Figure 5B) but reduced the recovery of PINP. Diluting 10 µL of plasma with 50 µL of ammonium carbonate followed by addition of either 60 or 80 µL of reduction/alkylation cocktail increased the recovery of PINP significantly (Figure 5B) with minimal loss (Figure 5C). Addition of 80 µL reduction/alkylation cocktail gave cleaner background than addition of 60 µL (data not shown), probably by precipitating more proteins without losing PINP. More than 90% of total protein was removed under this condition as visualized by Coomassie staining of the proteins after SDS-PAGE of the samples from reduction/ alkylation of the plasma compared to whole plasma (data not shown). Recovery yield of the PINP or the internal standard peptide did not differ between 75 or 80 µL of reduction/ alkylation reagent (Figure 5C). Thus the assay was tolerant to a small variation in the delivery volume of the reduction/ alkylation reagent which might result from the high vapor 4224

Journal of Proteome Research • Vol. 6, No. 11, 2007

Figure 5. Enrichment of PINP by ACN fractionation and optimization of recovery. Many proteins in the serum or plasma precipitate during reduction/alkylation due to high concentration of acetonitrile (ACN) whereas small peptides such as iSTDext peptide (23 AA residues) remain soluble at higher concentrations of ACN. iSTDext peptide was spiked into rat serum and recovery of pNTTP (originating from endogenous PINP) was compared to that of iSTD (originating from the iSTDext peptide). Peak intensity of the iSTD did not decrease as the ACN concentration was increased whereas that of pNTTP depended critically on ACN concentration. Volume of each component during reduction/ alkylation was optimized to achieve both maximum recovery of PINP and removal of unrelated proteins from the serum by comparing the recovery of PINP to that of iSTD. (A) Two microliters of rat plasma from commercial source was diluted to the indicated volume with 100 mM ammonium carbonate buffer (pH 11) and mixed with constant volume of reduction/ alkylation reagent prepared in ACN. Each sample contained the same amount of iSTDext peptide. (B) Rat plasma from a 8 monthold rat was diluted to a total volume of 60 µL by ammonium carbonate buffer, and different volumes of reduction/alkylation reagent prepared in ACN was added. (C) Optimum conditions for maximum recovery of PINP with clean signal are indicated for different volumes of rat plasma processed. Error bars indicate standard deviation of triplicate measurements. Signal intensities obtained for the first group were used as comparison to calculate recovery yield of the peptide under different conditions.

research articles

Rat PINP as a Biomarker for Bone Formation

pressure of acetonitrile. Up to 20 µL of serum or plasma could be handled in the same format simply by increasing volumes of other reagents proportionately (Figure 5C). The assay procedure was designed so that all sample-handling could be done in a 96-well PCR plate to increase throughput of the assay (see Experimental Section for details). Choice of Standard Dilution Matrix and Statistical Validation of the Assay. We evaluated sera from different animals as a diluent for the pNTTPext peptide to prepare calibration standards. We avoided solutions such as phosphate buffered saline (PBS) or BSA solution in PBS because these may not adequately mimic the complex environment of PINP in the serum or plasma during various steps of sample preparation and LC-MS/MS analysis. Fetal calf serum and guinea pig plasma appeared to degrade the synthetic peptides quickly even when the samples were kept at 4 °C (data not shown). Human, dog, sheep, pig, and rabbit sera had high levels of interfering noise (data not shown), limiting the working range of the assay to higher than 10 nM. Among the evaluated sera, horse serum and goat serum gave the cleanest backgrounds. Although the different dilution matrices gave different backgrounds at lower concentrations of the synthetic peptide, measured concentrations of PINP in test samples agreed well with each other at concentrations higher than 10 nM regardless of which serum was used as the diluent for the standard peptide, except for the fetal calf serum and the guinea pig serum (data not shown). Using synthetic standard peptides prepared in horse serum and goat serum, we performed a 3-day spike-recovery experiment to estimate the working range of the assay. Serial dilutions of the standard peptide (pNTTPext) were made in horse serum starting from 400 nM down to about 1 nM by making 3:4 dilutions at each step. Out of a total of 24 samples, even numbered samples were taken and designated as CalA and odd numbered samples were taken and designated as CalB. Thus, within each series, a ratio of concentrations between adjacent samples was 9/16, or roughly 2-fold different. We treated the CalA series as regular calibration standards, whereas the CalB series were treated as if they were unknown samples. Experiments were done in triplicates and concentrations of PINP in the CalB series were calculated using CalA as calibration standards which were compared to the known concentrations of these samples. Errors in measurement at each concentration were calculated to estimate accuracy and precision of the assay. This exercise was repeated three times on three different days. We then reversed the roles of CalA and CalB to predict concentrations of samples in the CalA series using CalB as the calibration set. Results of these two analyses were combined to obtain an overall assay working range, which was defined by the lowest and highest concentrations that satisfy the given criteria of assay precision (inter-batch % CV < 25% for lowest limit, 20% for other concentrations), assay accuracy (absolute value of % relative error < 25% for lowest limit, 20% for other concentrations) and overall assay performance (total error ) inter-assay % CV + absolute value of % relative error < 30% for all concentrations). The assay working range was from 1.27 nM to 300 nM. Average values for total error, absolute value of % relative error, and inter-batch % CV were 11.2, 2.9, and 8.3%, respectively, within the working range (Figure 6A and B first column). Maximum working range was limited by the maximum concentration that was tested, but may be extended higher than 300 nM because there was no sign of deterioration of the assay at high concentrations of the standard.

We also prepared two sets of calibration standards (CalC and CalD) in goat serum in a similar manner as in horse serum, and repeated the same experiment. The working range of the assay (2.26 to 300+ nM), overall assay precision, and accuracy obtained with these calibration standards were similar to those with calibration standards in horse serum (Figure 6B, fourth column). To test if different dilution matrices introduced any bias, we paired CalA from the horse serum with CalD from the goat serum and repeated the same analysis (Figure 6B, second column). We also tested pairing of CalB from the horse serum with CalC from the goat serum (Figure 6B, third column). All of these tests returned similar assay working ranges and assay precision, demonstrating that choice of dilution matrix did not introduce any systematic bias in determination of PINP concentration. PINP was stable in the rat serum when left overnight at 4 °C or at room temperature (Figure 6C). Repeated freeze-thawing of the serum did not affect PINP concentration in the serum (Figure 6C). Although there was a tendency for measured PINP concentration to increase slightly after multiple freeze-thaw cycles or overnight incubation, this observation was not statistically significant. Similar observations were made with rat plasma (data not shown). PINP as a Serum Biomarker of Bone Formation Activity. To ensure that the assay had adequate sensitivity to quantify PINP in rats of all ages, and to clarify age-dependent changes in PINP, we measured serum PINP from rats of different ages. Ages of these animals ranged from 3.5 weeks to 2 years old. Circulating levels of PINP showed a dramatic increase between 3.5 and 5 weeks, reaching a high of 260 nM. At about 9 months, PINP declined to only 13 nM. PINP continued to decrease, although at a slower rate, up to 2 yrs old (7.6 nM). The working range of the assay (1.27-300 nM) was sufficient to handle samples from rats of 1-24 months of age. To evaluate PINP as a serum biomarker for bone formation activity and to evaluate the usefulness of the assay, we treated adult rats with a bioactive form of parathyroid hormone (PTH), PTH 1-38, and prednisolone. PTH and prednisolone are known to modulate osteoblastic bone formation activity in opposite directions. PTH has been shown to increase bone formation in rats,2,12 whereas prednisolone was shown to retard bone formation in rats.18,19 To compare PINP to other frequently used markers of bone formation, we also measured bone alkaline phosphatase and osteocalcin from the same serum samples. After 10 days of daily treatment, PTH 1-38 increased PINP dose-dependently and significantly, whereas prednisolone strongly suppressed PINP (Figure 8A). Prednisolone inhibition was partially reversed by concomitant treatment with PTH 1-38. Therefore, PTH 1-38 and prednisolone affected PINP, as expected. Bone alkaline phosphatase (Figure 8B) also showed the same trend, but none of the changes reached statistical significance. Serum osteocalcin (Figure 8C) increased significantly with PTH 1-38 treatment in a dose-dependent manner similar to PINP, but the decrease in osteocalcin with prednisolone was not statistically significant. Overall, all three markers showed the same trend, but PINP showed more significant changes than the other two markers in both directions: decrease upon treatment with an inhibitor of bone formation activity and increase with a bone formation agent.

Discussion We have developed a mass spectrometry-based quantitative assay for rat PINP. Mass spectrometry is well established as a Journal of Proteome Research • Vol. 6, No. 11, 2007 4225

research articles

Han et al.

Figure 6. Statistical validation of the assay. (A and B) Three-day spike and recovery experiments were performed as described in the Results to assess variability of the assay. Two different dilution matrices were evaluated as carrier for standard peptide; in the CalA and CalB series, the standard peptide was diluted in the horse serum whereas goat serum was used as diluent for CalC and CalD series. (A) Results with CalA and CalB series. (B) Results when different combinations of the calibration series were used. Working range of the assay is the concentrations between LQL and UQL (lower and upper quantitation limit, respectively), which are defined according to the recommendation by the American Association of Pharmaceutical Scientists (AAPS) committee by the lowest and highest concentration that meet all of the following criteria; % bias (|% Rel Err|) < 20% (