Article pubs.acs.org/jpr
An Integrated Workflow for Multiplex CSF Proteomics and PeptidomicsIdentification of Candidate Cerebrospinal Fluid Biomarkers of Alzheimer’s Disease Mikko Höltta,̈ † Lennart Minthon,‡ Oskar Hansson,‡ Jessica Holmén-Larsson,† Ian Pike,§ Malcolm Ward,§ Karsten Kuhn,⊥ Ulla Rüetschi,† Henrik Zetterberg,†,¶ Kaj Blennow,† and Johan Gobom*,† †
Clinical Neurochemistry Laboratory, Institute of Neuroscience and Physiology, Department of Psychiatry and Neurochemistry, The Sahlgrenska Academy, University of Gothenburg, 431 80 Mölndal, Sweden ‡ Clinical Memory Research Unit, Department of Clinical Sciences Malmö, Lund University, 221 00 Lund, Sweden § Proteome Sciences PLC, KT11 3EP London, United Kingdom ⊥ Proteome Science R&D GmbH&CoKG, 60438 Frankfurt am Main, Germany ¶ UCL Institute of Neurology, Queen Square, WC1N 3BG London, United Kingdom S Supporting Information *
ABSTRACT: Many disease processes in the brain are reflected in the protein composition of the cerebrospinal fluid (CSF). In addition to proteins, CSF also contains a large number of endogenous peptides whose potential as disease biomarkers largely remains to be explored. We have developed a novel workflow in which multiplex isobaric labeling is used for simultaneous quantification of endogenous CSF peptides and proteins by liquid chromatography coupled with mass spectrometry. After the labeling of CSF samples, endogenous peptides are separated from proteins by ultrafiltration. The proteins retained on the filters are trypsinized, and the tryptic peptides are collected separately. We evaluated this technique in a comparative pilot study of CSF peptide and protein profiles in eight patients with Alzheimer’s disease (AD) and eight nondemented controls. We identified several differences between the AD and control group among endogenous peptides derived from proteins known to be associated with AD, including neurosecretory protein VGF (ratios AD/controls 0.45−0.81), integral membrane protein 2B (ratios AD/controls 0.72−0.84), and metallothionein-3 (ratios AD/controls 0.51−0.61). Analysis of tryptic peptides identified several proteins that were altered in the AD group, some of which have previously been reported as changed in AD, for example, VGF (ratio AD/controls 0.70). KEYWORDS: proteomics, peptidomics, Alzheimer’s disease, neurodegenerative disease, biomarker discovery, cerebrospinal fluid, isobaric labeling, quantification, clinical proteomics
■
disease.5,6 One of the main challenges with this approach is the broad scattering of the quantitative data caused both by individual variations in protein abundances and in analytical variation. Having a chance to detect statistically significant disease-related protein dysregulations necessitates the analysis of large cohorts. The ability to analyze large cohorts is practically limited by the long time required for liquid chromatography coupled with mass spectrometry (LC−MS). The multiplexing capacity of isobaric labeling techniques, such as the tandem mass tag (TMT) approach,7 currently enables up to 10-plexing, which decreases the number of analyses drastically and thereby enables larger studies to be performed. Increased multiplexing also decreases the incidence of missing data that occurs due to mass spectrometric undersampling when matching peptide identifications from multiple LC−MS data sets, which is a significant problem for the analysis of large
INTRODUCTION For most neurodegenerative and psychiatric diseases there are few or no biomarkers that can aid in diagnosis or be used for monitoring disease progression. For Alzheimer’s disease (AD), the most common form of dementia, three cerebrospinal fluid (CSF) biomarkers have been discovered, which are used to aid diagnosis: amyloid-β (Aβ) 1−42, total tau, and phosphorylated tau.1−3 While these markers have high diagnostic sensitivity and specificity for AD, there is a need to identify additional biomarkers that may allow making diagnosis at an earlier stage, which would enable monitoring of disease-modifying therapies and disease-related processes, to provide new insights into disease mechanisms. As CSF is in direct contact with the brain, the molecular composition of the CSF reflects many processes in the central nervous system,4 which makes it, arguably, the optimal fluid source of biomarkers for diseases of the central nervous system. Proteomic techniques based on mass spectrometry have been employed in several studies to identify new CSF biomarkers of © 2014 American Chemical Society
Received: June 17, 2014 Published: December 9, 2014 654
dx.doi.org/10.1021/pr501076j | J. Proteome Res. 2015, 14, 654−663
Journal of Proteome Research
Article
phosphorylation status, as previously described.22 Aβ 1−42 levels were determined using a sandwich ELISA (INNOTEST ß- AMYLOID(1−42), Innogenetics, Gent, Belgium), specifically constructed to measure the Aβ 1−42 peptide, based on combining the monoclonal antibodies 21F12 that are specific for the C-terminus and 3D6 that needs the first amino acid in the Aβ sequence.23 Deidentified CSF samples collected at the Clinical Neurochemistry Laboratory at the Sahlgrenska University Hospital/ Mölndal were pooled together and used for method development and evaluation of the assay precision. This procedure has been approved by the Ethics committee at the University of Gothenburg.
sample sets. Several proteomic studies in CSF have used proteolytic digestion followed by isobaric labeling for biomarker discovery in AD, Parkinson’s disease, and multiple sclerosis.8−13 While the majority of CSF proteomic studies to date follow the mainstay approach of digesting the sample proteins with trypsin, because of the high detection sensitivity and identification success rate for tryptic peptides, relatively little attention has been paid to endogenous CSF peptides. Naturally occurring peptides in CSF reflect a multitude of processes in the brain such as enzymatic activity, secretion, and aggregation. Endogenous peptides in the CSF have rarely been studied as potential biomarkers, although it has been shown that CSF contains many endogenous peptides of potential interest.14−19 The few quantitative studies that have been performed on endogenous peptides in CSF have employed label-free approaches to identify biomarkers for AD.15,17 These studies have identified some potential biomarker candidates and indicate that there are endogenous peptides of interest in CSF. We recently reported on a sample preparation based on ultrafiltration that enabled the identification of 730 endogenous CSF peptides14 by LC−MALDI MS. In the current work, we have modified this sample preparation and implemented TMTlabeling for quantification and on-filter tryptic digestion to enable analysis of both endogenous and tryptic peptides from the same CSF sample aliquot. We here demonstrate for the first time that isobaric labeling can be used to quantify endogenous peptides in CSF, which will pave the way for large-scale clinical peptidomic studies in CSF.
■
Study Design
The preparation of CSF samples was based on a previously described method,14 with additional steps for TMT labeling for quantification and on-filter tryptic digestion for concomitant proteome and peptidome analysis. The eight AD and eight control CSF samples (100 μL from each patient) were grouped into four TMT 6-plex sets, each set containing four patient samples, a CSF pool used to monitor the interset reproducibility, and a reference consisting of an equal-volume mixture of all samples in the entire sample set to enable comparisons between sets. Patient samples and the CSF pool were labeled using TMT reagents 126−130, while the reference sample was consistently labeled using TMT-131. Labeled samples within each TMT set were combined into one vial and processed according to the workflow outlined in Figure 1. Endogenous peptides were recovered in the flow-through after ultrafiltration. Proteins retained on the 30 kDa molecular weight cutoff (MWCO) filter were subjected to on-filter tryptic digestion, after which the tryptic peptides were recovered by a second step of centrifugation. Acquired MS/MS spectra were
EXPERIMENTAL SECTION
Samples
Clinical CSF was sampled according to a standard protocol4 at the Memory clinic in Malmö and consisted of a set of eight patients diagnosed with AD and eight nondemented controls, consisting of patients seeking medical advice due to memory problems but who did not fulfill the diagnostic criteria of AD. At the clinical baseline visit, physicians with special interest in cognitive disorders performed a thorough physical, neurological, and psychiatric examination as well as a detailed clinical interview focusing on cognitive symptoms and activities of daily living. The patients who received an AD diagnosis at baseline met the DSM-IIIR criteria of dementia20 and the criteria of probable AD defined by NINCDS-ADRDA.21 The demographics and clinical characteristics of the study groups are listed in Table 1. CSF total tau (T-tau) concentration was determined using a sandwich enzyme-linked immunosorbent assay (ELISA) (Innotest hTAU-Ag, Innogenetics, Gent, Belgium) specifically constructed to measure all tau isoforms irrespectively of Table 1. Demographics and Clinical Characteristics of the Study Groups N sex (F/M) agea Aβ 1−42 (pg/mL)a tau (pg/mL)a MMSEa a
control
AD
8 4/4 65 (51, 72) 700 (545, 818) 405 (330, 523) 29.5 (29, 30)
8 6/2 77 (72, 81) 235 (210, 298) 800 (717, 925) 24 (18, 26)
Figure 1. Workflow for TMT 6-plex labeling of endogenous peptides and proteins in CSF. CSF samples were subjected to reduction and alkylation of cysteine residues followed by TMT labeling and combining of the samples into TMT 6-plex sets. Ultrafiltration with 30 kDa MWCO filters was used to isolate the endogenous peptides from the proteins. The proteins retained on the MWCO filter were subjected to on-filter tryptic digestion, after which the tryptic peptides were collected by centrifugation. Following desalting on C18 cartridges, the endogenous and the tryptic peptides were analyzed by LC−MS.
Data are shown as medians with interquartile ranges. 655
dx.doi.org/10.1021/pr501076j | J. Proteome Res. 2015, 14, 654−663
Journal of Proteome Research
Article
Dionex nano-LC instrument (Ultimate 3000 RSLC, Thermo Scientific) fitted with a 75 μm × 2 cm trap column (PepMap Acclaim C18, Thermo Scientific) and a 75 μm × 50 cm separation column (PepMap Acclaim C18, Thermo Scientific) coupled to a Q-Exactive mass spectrometry instrument (Thermo Scientific). Peptide separation was performed using a 160 min gradient running from 3−45% of mobile phase B (84% ACN, 0.1% formic acid). The Q-Exactive was operated in the positive ion mode. The instrument settings for the MS scans were: resolution, 70 000; m/z range, 400−1600; max injection time, 250 ms; AGC target, 1e6. Data-dependent acquisition was used to record up to 10 consecutive fragment ion spectra (MS2) per full scan spectrum by selecting precursor ions in decreasing order of intensity and using 20 s dynamic exclusion and charge state exclusion to exclude signals with unassigned charge, charge 1 and >5. The isolation window was set to 1.2 m/z. The instrument settings for the MS2 scans were: resolution, 35 000 for endogenous peptides and 17 500 for tryptic peptides; fixed first mass m/z, 100; max injection time, 120 ms for endogenous peptides and 60 ms for tryptic peptides; and AGC target, 1e5. The data were searched using Mascot (Matrix sciences) against the human subset of the SwissProt database. The search settings were: oxidation of methionine as variable modification and TMT 6-plex (K), TMT 6-plex (N-term), and carbamidomethylation of cysteine as fixed modifications. For tryptic peptides, the enzyme specificity set to cleavage at arginine with one missed cleavage allowed. For endogenous peptides, no cleavage specificity was specified in the search, that is, the searches were performed against all possible peptide sequences within each protein sequence. For endogenous peptides, searches were also performed with acetyl (lysine) and pyroGlu (N-term) as fixed modifications instead of TMT 6-plex (K) and (N-term), respectively. The m/z error tolerances were 10 ppm in the MS mode and 20 milli-mass units in the MS/MS mode. The Percolator algorithm was used for peptide scoring. For identification, based on decoy searches, a target FDR value of 5% was used. TMT reporter ion ratios were calculated as the ratio of the respective TMT reporter ion (126−130) intensity to the reference TMT reporter ion 131 intensity. The relative difference in the abundances of endogenous peptides in AD, reported in Table 3, was calculated as the ratio of a peptide’s median TMT reporter ion ratios in the AD and the control group, respectively. For proteins, their relative abundances were calculated by taking the median of the median TMT reporter ion ratios of all the tryptic peptides identified from the given protein. As with the endogenous peptides, the relative abundance differences (reported in Table 4) were calculated as the ratios of the relative protein abundances in the AD and control group, respectively. MS/MS spectra with >30% coisolation (interference) in the precursor selection step were excluded from quantification. Two technical replicates of the endogenous peptides were analyzed by LC−ESI MS. Of all the quantified endogenous peptides in the AD study, 82% had a CV % of less than 20% between technical replicates.
used for peptide identification by database searching. The TMT reporter ion cluster in the low-mass region of the MS/MS spectra was used for quantification. The relative abundance of a given peptide in each individual sample was calculated relative to the reference by taking the intensity ratio of the corresponding reporter ion to the reference (TMT-131), which thus allowed comparison of the peptide’s abundance across all TMT sets. Sample Preparation
TMT Labeling. The CSF samples (100 μL) were mixed with 50 μL of 8 M guanidine hydrochloride solution (SigmaAldrich) and 15.6 μL of 1 M triethylammonium bicarbonate buffer (TEAB) (Sigma-Aldrich). Cysteine disulfides were reduced by adding 4.2 μL of 200 mM tris(2-carboxyethyl)phosphine (TCEP) (Thermo Scientific) and incubated at 55 °C for 1 h, after which the samples were allowed to cool to room temperature. For alkylation of cysteines, 4.3 μL of 400 mM iodoacetamide (Sigma) was added, followed by incubation at room temperature in the dark for 30 min. TMT 6-plex reagents (Thermo Scientific) were dissolved in acetonitrile (ACN) to a concentration of 19.5 mg/mL, and 19.5 μL of reagent was added to each sample, after which the samples were incubated for 1 h at room temperature. The TMT labeling reaction was then quenched by adding 9.7 μL of 5% (v/v) hydroxylamine (Sigma-Aldrich) to the samples, after which they were pooled into TMT sample sets. Isolation of Endogenous Peptides. MWCO filters (Vivacon 2 HY, Sartorius stedim), 30 kDa, were prewashed by spinning through 1 mL of 100 mM TEAB at 2500g at room temperature. The samples were then added to the filters that were centrifuged for 90 min at 2500g at room temperature. An additional volume of 200 μL of 100 mM TEAB was added to the filters and spun through to increase peptide recovery. Ultrapure water was added to decrease the ACN concentration to less than 5%, and the samples were acidified with 10% (v/v) trifluoroacetic acid (TFA). The samples were desalted on SEPPAK C18 cartridges (1 cm3, 100 mg, Waters), divided into three aliquots, dried in a speedvac, and stored at −80 °C pending analysis. On-Filter Digestion. Proteins retained on the 30 kDa filters were washed by spinning through 200 μL of 1% sodium deoxycholate (w/v, Sigma-Aldrich) in 50 mM ammonium bicarbonate buffer and then with 200 μL of 50 mM ammonium bicarbonate buffer. Trypsin (Sequencing grade, Promega) was reconstituted in 50 mM ammonium bicarbonate buffer to a concentration of 40 ng/μL, and 200 μL of the solution was added to each filter. The filters were incubated overnight at 37 °C. The tryptic peptides were then recovered from the filter by centrifugation at 2500g for 90 min, and an additional 200 μL of 50 mM ammonium bicarbonate buffer was spun through for increased recovery. The samples were then acidified with 10% TFA (v/v) and desalted, dried, and stored in the same way as the endogenous peptides. LC−ESI MS
Both endogenous and tryptic peptides were analyzed by ESI− MS. Lyophilized sample aliquots corresponding to 1/3 of the original sample were reconstituted in 2% ACN, 0.1% TFA (v/ v). Extracts of endogenous peptides were reconstituted in 6 μL, corresponding to a total volume of 200 μL of CSF, of which all was loaded on the LC. Tryptic peptide samples were dissolved in 200 μL, and 4 μL, corresponding to the same volume of CSF, was loaded on the LC. LC−MS was performed on a
LC−MALDI MS
The endogenous peptides were also analyzed by LC−MALDI MS according to a previously described method.14 Briefly, the peptides were separated by nano-LC using an identical LC setup to the one that was used for LC−ESI−MS. The peptides were fractionated onto MALDI plates using a microdispensing 656
dx.doi.org/10.1021/pr501076j | J. Proteome Res. 2015, 14, 654−663
Journal of Proteome Research
Article
Table 2. Accuracy and Precision of TMT Labeling in CSFa
robot (iTWO, M2 Automation, Berlin) set to dispense a droplet of the LC effluent onto the plate every 10 s. Matrix solution was then dispensed on the dried sample spots, and the MALDI plate with the samples was analyzed using a MALDITOF/TOF MS (UltrafleXtreme, Bruker Daltonics). Mass spectra were recorded in the positive ion mode over the m/z range of 700−5000 using deflection to suppress ions up to m/z 600. In MS-mode, each recorded spectrum was the sum of 2000 single-shot spectra recorded from 10 random positions on each sample spot. Using the Warp-LC software (Bruker Daltonics), compounds with S/N > 15 were selected for MS/ MS analysis. For MS2, 2000 single-shot spectra were accumulated of the precursor ion, followed by 4000 of the fragment ions. The data were searched using Mascot, as described in the previous section, with the differences that the m/z error tolerance was 15 ppm in the MS mode and 0.8 Da in the MS2 mode. Mascot’s score identity threshold was used as criterion for peptide identification, and the annotated MS2 spectra of all significantly altered peptides were manually evaluated. TMT reporter ion ratios were calculated relative to the TMT-131 reporter ion signals using the ProteomeScape software (Bruker Daltonics).
TMT reporter ion ratio TMT reporters
endogenous peptides
126/131 127/131 128/131 129/131 130/131
1.002 1.066 1.095 1.085 1.075
(0.094) (0.096) (0.098) (0.097) (0.104)
tryptic peptides 0.998 1.059 1.089 1.066 1.078
(0.113) (0.125) (0.134) (0.112) (0.132)
a
Aliquots of the same CSF pool were analyzed in a TMT 6-plex experiment. The data shown in the table are mean values and standard deviations for quantified endogenous and tryptic peptides based on ESI−MS data.
digestion with trypsin, and the produced tryptic peptides are recovered separately by centrifugation.24 While the N-hydroxy succinimide-activated reagents used in TMT and iTRAQ can be used to label intact proteins as well as proteolytic peptides, the former approach has been used in only a few studies.25−27 In clinical proteomics of biofluids, labeling on the protein level can be beneficial in that protein fractionation, for example, by SDS-PAGE or ion exchange chromatography, can be used to reduce the suppressive influence of high-abundant proteins, as was demonstrated in a proteomic study in serum.26 In contrast to serum and plasma, the significantly lower protein concentration in CSF (300−500 μg/mL) renders it possible to label proteins in about 300 μL of sample using the standard quantity of reagent (0.8 mg) without the need for first depleting high-abundant proteins. An advantage of performing the labeling reaction as in the first step in the sample preparation is that the risk of introducing errors due to varying recovery is avoided.
APL1β ELISA
The CSF samples from the clinical cohort were analyzed for the amyloid-like protein 1 (APLP1) peptides APL1β25, APL1β27, and APL1β28 using a commercial ELISA (IBL International, Hamburg). The samples were analyzed according to the kit insert with minor modifications. The CSF samples were diluted 1:20 for APL1β25, 1:10 for APL1β27, and 1:5 for APL1β28. All samples were analyzed in duplicate, and the CV% for standards and samples was
