Subjects and clinical samples
CSF samples involved in our study included a coded cohort of 12 untreated SPMS patients and untreated 12 non-inflammatory neurological disorders (NIND) controls from the Neuroimmunology Branch of the National Institutes of Neurological Disorders and Stroke (NIB/NINDS/NIH) collected under natural history protocol 09-N-0032, healthy controls from University of Hawaii, and SPMS patients from a UK double blinded lamotrigine trial (active arm and placebo). All patients signed informed consent and all research procedures were approved by the institutional review boards (IRB) of afore-mentioned institutions.
Cerebrospinal fluid sample preparation
NIB samples were transported to the laboratory on ice and spun (3000 g x 10 min) within 30 min of collection. Cell-free supernatants were sequentially coded and immediately cryopreserved at −80°C in 500 μl aliquots.
Prior to all sample preparation, we prepared a stock of 1pmol/μL of horse apomyoglobin in deionized water to be used as an internal standard spiked into all CSF samples. Prior to tryptic digestion of the CSF samples used both in method development and then in our cohort screening experiments, 100 μL of 0.1% Rapigest (Waters, Milford, MA) resuspended in 100 mM ammonium bicarbonate, 2 μL 500 mM dithiothreitol (DTT), and 10 μL of the 1pmol/μL stock of the internal standard, horse apomyoglobin, were added to each individual 100 μL CSF aliquot. To reduce protein disulfide bonding, each sample aliquot was heated at 60°C for 30 min. After allowing each sample cool down to room temperature, 10 μL of 200 mM iodoacetamide (IA) was added. Each vial was placed in the dark for 30 min to allow for alkylation of all free protein cysteine residues. 5 μL of trypsin (0.2 μg/μL in 1 mM HCl) was then added to each sample following the DTT reduction and IA alkylation steps. The sample mixtures were incubated at 37°C for 18 h to allow for complete enzymatic digestion of the CSF and the protein standard horse apomyoglobin. Following trypsinization, 2 μL of trifluoroacetic acid was added to the digestion mixture to quench the reaction. The samples were then incubated at 37°C for an additional 45 min. Following this second incubation step, samples were speed-vacuumed to dryness, reconstituted in 0.1% formic acid in deionized water, and desalted using Waters Oasis HLB solid phase extraction cartridges (Waters, Milford, MA) according to the manufacturer’s protocol with a vacuum manifold. To prepare for LC-MS analysis, the eluents from the solid phase extraction were speed-vacuumed to dryness and reconstituted with 20 μL 0.1% formic acid in water. Pooled CSF samples obtained from patients without identified medical records were used during the initial stages of method development. All reagents were added to practice CSF in proportion as described. Diseased CSF and healthy control CSF samples selected for inclusion for our cohort screening experiments were obtained from patients with complete associated medical records. Individual CSF patient samples used during cohort screening were blinded prior to any sample processing. Additionally, sample aliquots were prepared in parallel and as duplicates to minimize experimental error during sample preparation. All reagents were purchased from Sigma if not mentioned specifically.
HPLC separations
Prior to our selected reaction monitoring assay development efforts, the RP-HPLC gradient profile used for separation of digested CSF samples was first optimized. An Agilent ZORBAX SB-C18 column (150 X 0.5 mm, 5 μm) (Agilent Technologies Inc., Santa Clara, CA, USA) was attached to a Waters Nano-Acquity ultra-high pressure HPLC (Waters, Milford, MA, USA) and placed in the front end of our Thermo Vantage Triple Quadrupole mass spectrometer (Thermo Fisher Scientific, Waltham, MA). Total Ion Chromatograms of the various c18 RP-HPLC separations were acquired in full scan mode acquisition with a scan time of 0.5 s. Ultimately, we selected a 32 min RP-HPLC linear gradient to be used for all subsequent CSF sample separation whereby the injected peptides were eluted with the following gradient: the first 2.4 min post-injection were diverted to waste with a 3% B hold, a linear 3-45% B gradient was performed in the next 22.6 min, a linear 45-90% B gradient ramp was performed in 1 min followed by a 90-100% B linear ramp in 1 min, 100% B was held for 1 min before a linear gradient of 100-3% B was performed in 2 min followed finally by a 3% B hold for 2 min. Solvent A consisted of 0.1% formic acid in water and solvent B consisted of 90% acetonitrile, 10% water and 0.1% formic acid. The flow rate of the Nano-Acquity HPLC was set at 20 μL/min.
LC-MS/MS method
Intact horse apomyoglobin was chosen as a model protein to establish the LOD of our multiplex SRM LC-MS/MS method. Digested horse apomyoglobin was diluted in serial in digested CSF with concentrations ranging from 1 fmol/μL to 100 fmol/μL. Using a CSF sample with 100 fmol/μL of digested horse apomyoglobin, the following was experimentally determined: the signature tryptic peptide NDIAAK was selected as the surrogate peptide for the apomyoglobin protein, the three most abundant detected transitions of peptide NDIAAK, ions y3, y4, and y5 were selected for detection in Q3, and of those three, the most abundant, y3, was selected for generating a standard curve for this peptide. A serial of apomyoglobin dilutions were then subject to LC-MS/MS analysis with Q1 mass filter targeting detection of the parent ion of NDIAAK and Q3 mass filter targeting detection of the three selected product ions. Ions were introduced into the mass spectrometer via an H-ESI II probe outfitted with a 32 gauge needle (Thermo Fisher Scientific, Waltham, MA, USA). Instrument operating parameters used during acquisition were as follows: capillary temperature of 270°C, vaporizing heat was shut off, sheath gas pressure was set at 10, no auxiliary gas was used, and the spray voltage was set to 4,000 V. Instrument settings in Xcaliber were as follows: positive scan mode, scan widths of 0.004 m/z, scan times of 0.015 s, chrom filters enabled and set at 50, collision gas set at 1.5mTorr, Q1 peak width (FWHM) of 0.7, Q3 peak width of (FWHM) 0.7, and cycle times of 5 s (Xcaliber v 2.1, Thermo Fisher Scientific, Waltham, MA, USA). Following data acquisition, peak areas under the curve were calculated using an open source proteomic software platform Skyline [23] (Skyline v. 1.1, MacCoss Lab, Seattle, WA) and a linear standard curve (y = 5884.7x-9111.7, R2 = 0.9998) was generated using Microsoft Excel (Figure 1).
LC-MS/MS method development and optimization
Using a Waters ultra-high pressure reverse phase high performance liquid chromatography (Waters nanoACQUITY UPLC, Milford, MA, USA) outfitted with an Agilent ZORBAX SB-C18 column (150 X 0.5 mm, 5 μm) (Agilent Technologies Inc., Santa Clara, CA, USA) and coupled to a Thermo Fisher triple quadrupole mass spectrometer (Thermo TSQ Vantage, Thermo Fisher Scientific, Waltham, MA, USA), we designed a multiplexed peptide based selected reaction monitoring (SRM) assay using the RP-HPLC and mass spectrometric settings mentioned in the two previous method sections. Our SRM assay allows for the simultaneous relative quantification of 26 tryptic peptides originating from proteins implicated in SPMS pathology. FASTA files for selected protein biomarkers were downloaded from Uniprot (http://www.uniprot.org) individually and imported into Skyline. The uniqueness of the selected surrogate peptides was confirmed by running individual BLAST search from NCBI website (http://blast.ncbi.nlm.nih.gov/Blast.cgi). Instrumental methods were exported from Skyline into a Thermo Xcaliber 2.1 “EZ SRM” method template with the same settings as used during data acquisition for our LOD determination experiments (Xcaliber v. 2.1, Thermo Fisher Scientific, Milford, MA, USA). To design our SRM assay, we adopted an iterative method development strategy, eliminating poor performing transitions from the SRM assay in a targeted method refinement cycle [23]. Raw data was inspected in the Skyline software platform and the peak areas of each detected transition were normalized to that of horse apomyoglobin peptide HGTVVLTALGGILK. Starting with an initial 1885 transitions for 403 surrogate peptides, we completed seven rounds of iterative method refinement to generate the final SRM assay. The final SRM assay was a scheduled method consisting of 78 transitions for 26 peptides with a scheduling window of four minutes - no more than 50 concurrent transitions were detected at any given time. Reproducibility was confirmed by running four replicates. Both retention time and peak area are highly reproducible (Figure 3a and 3b).
Data analysis
Box-Cox transformation was applied to CSF values of the 26 biomarkers. Age and gender were considered as covariates, p-value of 0.1 was used for covariate selection. Analysis of variance (ANOVA) with or without covariates was performed to assess the differences in the CFS biomarkers values between three cohorts (22 SPMS, 12 NIND and 10 HC), followed by Scheffe’s pair-wise comparison of means. ANOVA was also applied to assess the effect of lamotrigine therapy on the expression of 26 protein biomarkers by comparing 22 SPMS patients with 7 lamotrigine treated SPMS patients.
The relationship between the 26 biomarkers was estimated by Pearson correlation coefficients. Since these biomarkers were found to be closely correlated, principal component analysis (PCA) with Varimax rotation was performed to represent the 26 biomarkers as a set of new orthogonal variables, and subsequently, ANOVA with or without covariates was applied to the selected principal components.
A significant level of 0.01 was used in order to adjust for multiple testing. The statistical analyses were performed using SAS version 9.2.