Identification of POTENTIAL Cerebrospinal Fluid

Identification of POTENTIAL Cerebrospinal Fluid

1

Online Supplemental Data

Identification of POTENTIAL cerebrospinal fluid

biomarkers IN amyotrophic lateral sclerosis

Pasinetti G.M., MD, PhD;Ungar L.H., PhD; Lange D.J., MD;Yemul S., PhD;

Deng H., PhD; Yuan, X., PhD;Brown R.H., MD, DPhil; CudkowiczM.E., MD;

Newhall K., BA; Peskind E., PhD; Marcus S., PhD and Ho L., PhD

methods

Subjects

Two studies were performed on CSF from patients with ALS identified from the ALS programs at the Mt Sinai School of Medicine, Massachusetts GeneralHospital, and University of Washington. In a first discovery study, CSF from normal subjects (n=21) and patients with ALS (n=36) were used for the SELDI-MS analysis. Patients with ALS were classified as having either definite or probable ALS according to the WFN El Escorial diagnostic criteria (1) (Table 1); CSF specimens from control subjects were obtained from compensated community volunteers in good health (2). In a second study for validation, SELDI-MS analysis was performed on CSF from a second set of normal subjects (n=25), a different cohort of ALS (definite or probable) patients (n=13) and patients with peripheral neuropathy (n=7; 5 with multifocal motor neuropathy with conduction block, 2 with sensorimotor peripheral neuropathy of uncertain cause).

All CSF samples used were derived from comparable fractions (e.g. 20–25 ml), to limit variability from rostro-caudal concentration gradients. Following collection, samples were gently mixed, divided into aliquots and immediately frozen in dry-ice and stored at –80oC. Human Subjects Division of the participant institutions (MSSM, MGH, and UW) approved the collection of samples and written informed consent was obtained from all subjects.

Proteomic technology

SELDI-MS is a system that enables rapid protein profiling, identification, and characterization from crude biological samples by selectively capture subclasses of proteins with specific physical or biochemical characteristics. The molecular size (MW) as well as the quantity of individual proteins absorbed on each chip is then directly assessed by a “time-of-flight” mass spectrometer, generating quantitative protein mass profiles for individual CSF specimen. Comparative protein expression profile analysis highlights CSF protein species which are aberrantly regulated in the CSF of ALS compared to control subjects.

Before analysis of the CSF, experiments were designed to optimize the ProteinChip type and washing conditions for the SELDI-MS proteomics tests. A total of three chip types, SAX2 (strong anion exchange surface), WCX2 (weak cation exchange surface), and IMAC3 (metal binding surface) (Ciphergen Biosystems, Inc., Palo Alto, CA) and varying binding/washing conditions, including pH 4, 6, 8, and 10 binding/washing buffers were tested. The ProteinChip type and wash condition that gave optimal results was the WCX2 ProteinChip with 100 mM ammonium acetate, pH 4 with 0.1% Triton X-100 for equilibration, binding and wash steps. This was therefore used in this study.

The WCX2 ProteinChip with 100 mM ammonium acetate, pH 4 with 0.1% Triton X-100 for equilibration, binding and wash steps was used for SELDI-MS analysis of CSF specimens. For all experiments, a 96-well bio-processor was used. ProteinChip spot surface was pre-treated with 250 µL 10 mM HCl for 10 minutes and then equilibrated with wash buffer (100 mM ammonium acetate, pH 4 with 0.1% Triton X-100) for 5 minutes. Fresh aliquots of CSF specimens stored at –80 oC were slowly thawed in ice, centrifuged (3000 x g, 4o C, 15 min) and equal amounts of CSF protein (1.5 µg) were applied to each ProteinChip spot and incubated in a humidity chamber at room temperature for 60 minutes with shaking at 250 rpm. Unbound proteins were removed by washing twice for 5 minutes with binding buffer and then briefly washed with HPLC grade water. An energy absorbing molecule (EAM), 20% sinapinic acid (SPA) in a solution containing 50% acetonitrile / 5% trifluoroacetic acid, was applied onto the ProteinChip spot surface. The ProteinChips were air dried and proteins captured on individual ProteinChip spots were evaluated using the PBSII ProteinChip reader (Ciphergen Biosystems, Inc.). The ProteinChip Reader was externally calibrated with the Ciphergen All-In-One peptide/protein standard mix (Ciphergen Biosystems, Inc., Palo Alto, CA) containing hirudin (7.03 kDa), bovine cytochrome-C (12.23 kDa), equine myoglobin (16.95 kDa), bovine red blood cell carbonic anhydrase (29.02 kDa), enolase (45.67 kDa), bovine albumin (66.43 kDa) and bovine IgG (147.30 kDa). The external calibration provided a 0.1% mass accuracy. The ionized proteins were detected and their molecular mass/charge (M/Z) ratios determined using Time-of-flight mass spectrometry (TOF-MS) analysis. Protein peaks were analyzed with the Ciphergen ProteinChip software version 3.0, as previously described (3). Each study was repeated at least twice and baseline subtraction, spectrum normalization, and peak detection were performed with the ProteinChip Software (Ciphergen Biosystems, Inc., Palo Alto, CA).

Purification and sequence analysis of the 13.4 kDa protein species

200 L CSF samples from a healthy control subject and a patient were reduced with 50 mM DTT and alkylated with 100 mM iodoacetamide. The Amicon ultracentrifuge filter (50,000 MWCO; Millipore) was used to pre-fractionate the sample to two fractions. The fraction that contained <50 kDa CSF protein/peptides was injected into a reversed phase C4 column and 20 fractions were collected. Proteins in the selected LC fractions based on the UV absorbance were further separated on 1-dimensional SDS PAGE and visualized by silver staining. In fraction 20, we identified a band at MW approximately 13-14 kDa showing preferentially higher contents in the control subject compared to the ALS patient (figure 1A). This gel band was excised from the gels and de-stained. The protein was digested in-gel with typsin for 16 h at 30 °C to generate tryptic peptide mixtures and the digestion was stopped by addition of 0.1% of TFA.

The tryptic peptide mixture was injected on a 0.3 mm (id 65 mm) trapping column (PepMap; Dionex) at a flow rate of 20 µL/min (total loading time of 5 min). By switching the stream valve in the SWITCHOS (Dionex), the trapping column is back-flushed with a binary solvent gradient, which is started simultaneously with the injection cycle. The sample is thereby loaded onto a capillary RP C18 column (0.2 id 50 mm Magic from Michrom BioResources, Inc.). Peptides were eluted from the stationary phase using a linear gradient from 0 to 75% solvent B (100% acetonitrile (ACN) in 0.1% formic acid) applied over a period of 45 min. The solvent delivery system was set at a constant flow rate of 20 µL/min. Using a 1/100 flow splitter, 2 L/min of solvent was directed through the capillary column. The outlet of the capillary column was in-line connected with a distal metal-coated fused silica PicoTip™ needle (PicoTip™ FS360-20-10-D-C7, New Objective, Woburn, MA, USA) that is interfaced with a ThermoFinnigan LCQ mass spectrometer or a QSTAR XL QqTOF mass spectrometer (Applied Biosystems).

Automated data-dependent acquisition was initiated after the stream valve was switched. The acquisition parameters were set so only doubly and triply charged ions were selected for fragmentation. The acquired CID-spectra were converted to a MASCOT acceptable format. The allowed variable modifications for database searching were oxidation of methionines. Only MS/MS spectra that were identified by a score that exceeded the identity threshold score of MASCOT at the 95% confidence level were retained. The retained spectra were subsequently validated manually. Only spectra that held a high number of typical fragment ions were considered to be positive (typically about 50% of b- and y- ions were present). The identified peptides were automatically stored in our database and links were made to their MS/MS spectra and precursor proteins.

Purification and sequence analysis of the 4.8 kDa protein species

1.8-ml CSF from an ALS and a control case was treated with anti-albumin IgY (anti-HAS kit, VivaScience, Hannover, Germany) to remove contents of albumin following manufacturer’s instruction. Subsequent to albumin depletion, samples were concentrated by vacuum centrifugation, and were fractionated with a reversed-phase C4 column. Protein species recovered in each fraction was analyzed by MALDI-TOF mass spectrometry. The 4.8 kDa ALS biomarker protein species was recovered in fraction 17.

Protein species in fraction 17 was desalted and concentrated using micro-C18 ZipTip (Millipore) following manufacturer’s instructions. Purified protein species from fraction 17 was recovered from the ZipTip with 2 l of an aqueous ACN solution (containing 50% ACN (v/v) and 0.1% formic acid) and load onto a GlassTip (New Objective; source, city) in preparation for quadrupole time-of-flight (Q-TOF) MS analysis. Q-TOF analysis was performed using an Applied Biosystems QSTAR Qq-TOF mass spectrometer operated in positive modes. The ions corresponding to the quadrupole charged 4808.8 species was selected for subsequent collision dissociation and resultant fragments ions were analyzed. Resultant MS/MS spectrum information was submitted to Mascot search tool for identification.

Quantitative assessment of cystatin C by ELISA

Cystatin C content in CSF was quantified using a commercial ELISA kit (America Laboratory Products). CSF samples were diluted by 400- and 800-fold in a dilution buffer provided by the EELISA kit and 100 l of the diluted CSF was assayed following manufacturer’s instruction. Cystatin C content was calculated based on a standard curve generated from purified cystatin C (provided by the kit).

Statistical Methods

Protein species peaks were tested for significant difference between groups using a two-sided heteroscedastic t-test. Receiver operating characteristic (ROC) curve was used to relate “sensitivity” and “specificity” or sensitivity at a given specificity for providing cut-off values (4-5). Statistical program package of social science (SPSS) 11.0 was used for making ROC curves and correlation analysis. P values for the difference between areas under the ROC curves were calculated using command ROCCOMP in statistical software Stata/SE 8.0 for correlated samples. The 95% confidence intervals for the difference between areas under the ROC curves were calculated using BOOTSTRAP method in statistical software S-Plus 6.0 for Windows. In this study, the “sensitivity” (true positive (TP)/(TP+ false negative (FN)) is the probability that a patient who was predicted to have ALS actually has it while the “specificity” (true negative (TN)/(false positive (FP)+TN) measures the probability that a patient predicted not to have ALS will, in fact, not have it.

FIGURE LEGENDS

Figure 1 Purification and sequence identification of the 13.4 kDa ALS biomarker species as cystatin C.

200 L CSF samples from a healthy control subject and an ALS patient were fractionated by reversed phase chromatography followed by SDS-PAGE. (A) Identification of a 13.4 kDa protein (indicated by an arrow), which is down-regulated in the CSF of ALS compared to control case. The 13.4 kDa protein species was digested with trypsin and three most abundant tryptic peptides were selected for LC-MS/MS analysis. Abbrev. MW, molecular weight marker; Ctl, control case; AL, ALS case. (B) LC-MS identification of one of the three major tryptic peptide species (indicated by arrow). Inset: The aa sequence of this peptide identified based on LC-MS/MS analysis. (C) Summary of the aa sequences of each of the three tryptic peptide subjected to LC-MS/MS analysis. (D) All three tryptic peptide derived from the purified 13.4 kDa protein species matched perfectly with the human cystatin C. Sequences which the three sequenced tryptic peptides matched with cystatin C are highlighted by an underline.

Figure 2 Purification and sequence identification of the 4.8 kDa ALS biomarker species as a proteolytic fragment of the neurosecretory protein, VGF.

1.8-ml CSF from an ALS and a control case was treated with anti-albumin IgY (anti-HAS kit, VivaScience, Hannover, Germany) to remove contents of albumin, followed by fractionation using reversed phase HPLC. (A) MALDI-TOF mass spectrometry identification of a select HPLC fraction containing enriched content of the 4.8 kDa protein (indicated by arrow). (B) Proteins recovered in this HPLC fraction were further analyzed by Qq-TOF. The 4.8 kDa protein species was identified as peak 1202.45 (indicated by an arrow), which is the expected molecular size of the monoisotopic 4804.8 Da mass species presented as a quadrupole charged species. Inset: presentation of the 4804.8 protein species as a quadrupole charged species. (C,D) The ions corresponding to the quadrupole charged 4808.8 species was selected for subsequent collision dissociation and resultant fragment ions were analyzed. The ESI-mass spectrums acquired for the major peaks identified after collision dissociation are assigned to either b-type (C) or y-type (D) ions. Based on the this information, the MS/MS spectrum was submitted to Mascot search tool for identification, which derived the peptide sequence:A(Q/K)NA(I/L)(I/L)FAEEED. By blast searching, we found only one protein, the neuro-endocrine specific protein VGF, whose sequence matched with the 4.8 kDa peptide sequence. (E) Protein sequence of the full-length precursor VGF polypeptide. The first 22 bases represent signal peptide (highlighted by a red-underline). Residues 23 to 616 represent the mature VGF protein. The sequence identified from the 4.8 kDa ALS biomarker march with a peptide fragment corresponding to aa 398-411 (highlighted by a red box) of the precursor VGF.

Figure 3 ELISA confirmation of reduce contents of cystatin C in the CSF of ALS compared to control cases.

CSF samples were diluted by 400- and 800-fold. 100 l of the diluted CSF was assayed using a commercial cystatin C ELISA (America Laboratory Products) following manufacturer’s instruction. Bar graphs represent mean + SD content of a cystatin C in the CSF of healthy controls (n=21) and ALS (n=36) cases; *p<0.0001 vs. controls by two-tailed t-test.

REFERENCES

1.Shefner JM, Cudkowicz ME, Schoenfeld D, Conrad T, et al. A clinical trial of creatine in ALS. Neurology 2004; 63:1656-1661.

2.Zhang J, Goodlett DR, Quinn JF, Peskind E, et al. Quantitative proteomics of cerebrospinal fluid from patients with Alzheimer disease. J. Alzheimer’s Dis 2005; 7:125-33.

3.Luo X, Carlson KA, Wojna V, et al.Macrophage proteomic fingerprinting predicts HIV-1-associated cognitive impairment. Neurology 2003; 60:1931-1937.

4.Baker SG. The central role of receiver operating characteristic (ROC) curves in evaluating tests for the early detection of cancer. J Natl Cancer Inst 2003; 95:511-515.

5.Pepe MS. An interpretation for the ROC cure and inference using GLM procedures. Biometrics 2000; 56:352-9.