A Combined Shotgun and Targeted Mass ... - ACS Publications

Subscriber access provided by NEW YORK UNIV

Article

A Combined Shotgun and Targeted Mass Spectrometry Strategy for Breast Cancer Biomarker Discovery Martin Sjöström, Reto Ossola, Thomas Breslin, Oliver Rinner, Lars Malmström, Alexander Schmidt, Ruedi Aebersold, Johan Malmström, and Emma Niméus J. Proteome Res., Just Accepted Manuscript • DOI: 10.1021/acs.jproteome.5b00315 • Publication Date (Web): 06 May 2015 Downloaded from http://pubs.acs.org on May 25, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Proteome Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 45

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Proteome Research

A Combined Shotgun and Targeted Mass Spectrometry Strategy for Breast Cancer Biomarker Discovery Martin Sjöström†*, Reto Ossola¶‡, Thomas Breslin†, Oliver Rinner‡, Lars Malmström±, Alexander Schmidt¤, Ruedi Aebersold&ǁ, Johan Malmström┴§, Emma Niméus†*# †

Division of Oncology and Pathology, Department of Clinical Sciences Lund, Lund University,

Lund, Sweden ‡

Biognosys AG, Schlieren, Switzerland

±

¤

ǁ

S3IT, University of Zurich, Zurich, Switzerland

Biozentrum, University of Basel

Department of Biology, Institute of Molecular Systems Biology, Eidgenössische Technische

Hochschule, Zurich, Switzerland &

Faculty of Science, University of Zurich, Switzerland

┴

Division of Infection Medicine, Department of Clinical Sciences Lund, Lund University, Lund,

Sweden §

Division of Surgery, Skåne University hospital, Lund, Sweden

ACS Paragon Plus Environment

1


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 45

KEYWORDS: breast cancer, biomarker, shotgun proteomics, targeted proteomics, LC-MS/MS, SRM, MRM, N-glycosylation, estrogen receptor, HER2.

ABSTRACT It is of highest importance to find proteins responsible for breast cancer dissemination, for use as biomarkers or treatment targets. We established and performed a combined non-targeted LCMS/MS and a targeted LC-SRM workflow for discovery and validation of protein biomarkers. 80 breast tumors, stratified for estrogen receptor status and development of distant recurrence (DR+/-) were collected. After enrichment of N-glycosylated peptides, label-free LC-MS/MS was performed on each individual tumor in triplicate. In total, 1515 glycopeptides from 778 proteins were identified and used to create a map of the breast cancer N-glycosylated proteome. Based on this specific proteome map, we constructed a 92-plex targeted label-free LC-SRM panel. These proteins were quantified across samples by LC-SRM, resulting in 10 proteins consistently differentially regulated between DR+/DR- tumors. Five proteins were further validated in a separate cohort as prognostic biomarkers at the gene expression level. We also compared the LCSRM results to clinically reported HER2 status, demonstrating its clinical accuracy. In conclusion, we demonstrate a combined mass spectrometry strategy, at large scale on clinical samples, leading to the identification and validation of five proteins as potential biomarkers for breast cancer recurrence. All MS data are available via ProteomeXchange and PASSEL with identifiers PXD001685 and PASS00643.


2

Page 3 of 45

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


MAIN TEXT Introduction Breast cancer is the most common female cancer in the world, estimated to cause 1.4 million new cancers each year.1 It is a heterogeneous disease2 showing distinct clinical behavior, treatment response and prognosis, depending on tumor characteristics.3 Pathological evaluation includes estrogen receptor alpha (ER) and human epidermal growth factor receptor 2 (HER2). ER status has a large impact on tumor behavior and determines the response to endocrine treatment.4,5 Overexpression or amplification of HER2 affects prognosis and determines the response to the HER2 directed therapy.6 Breast cancer prognosis is generally good, with five year survival rates approaching 90% in developed countries.1 However, metastatic disease is not curable and can occur several years after surgery. It is a profound challenge to predict recurring patients early and offer adjuvant treatment. New gene-profiling techniques have been developed to better evaluate the risk of recurrence7, and studies are ongoing to evaluate the clinical utility of these tests.8,9 However, analyzing breast tumors at the mRNA level does not provide a complete picture of cellular functions, as mRNA levels do not necessarily correlate with protein levels10-12 and proteins may be subject to post-translational modifications.13 This stresses the need to analyze the breast cancer proteome. Recently, proteomic profiles have been reported for grading and prognosis in breast cancer.14,15 Of particular importance is the analysis of sub proteomes relevant for recurrences because they are expected to indicate protein patterns that are not apparent from the corresponding DNA or RNA sequence. The N-glycosylated proteins constitute such an important sub proteome in breast cancer, as many of these proteins are normally present in the plasma membrane, the extracellular-matrix or are secreted to the extra cellular space. The functional protein groups


3


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 45

enriched in the N-glycosylated sub proteome are known to mediate cellular processes important to cancer progression such as cell-to-cell signaling, growth, differentiation and migration.16,17 In addition, many of these proteins have a higher propensity to be secreted into and to persist in the bloodstream. As technical issues have been resolved, mass spectrometry (MS) have become the method of choice for proteomic studies in breast cancer.18 Liquid-chromatography tandem MS (LCMS/MS), or “shotgun” MS, is used to identify up to thousands of proteins in samples of unknown protein composition. LC-MS/MS can also provide estimates of protein quantity, but the data dependent selection of precursor ions results in fragmented datasets with relatively large numbers of missing values if multiple samples are analyzed and compared. Further, quantification comes with a large confidence interval, especially when using label-free quantification strategies.19 Several liquid-chromatography tandem MS (LC-MS/MS) studies have described a discovery phase for biomarker discovery in breast cancer20-24 and shown promise in detecting relevant proteins. However, no marker derived from such studies has so far been introduced to clinical practice. This is largely due to lack of validation of the initial LC-MS/MS results. It has been proposed to combine LC-MS/MS discovery with complementary validation techniques.25-27 LC-coupled selected reaction monitoring (LC-SRM) is a MS technique recently adapted to the field of proteomics.28 It is based on previous knowledge of a peptide and gives good quantitative performance for a limited number of analytes; typically, peptides from tens of proteins can be analyzed in parallel. LC-SRM offers the opportunity to validate multiple biomarker candidates simultaneously, once the target proteins are known. Thus, it is tempting to use a strategy combining LC-MS/MS for a discovery phase and LC-SRM for targeted


4

Page 5 of 45

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


verification of discovered protein sets. Indeed, the approach has recently been described in breast cancer by several investigators.29-36 In this study, we applied a combined discovery and verification mass spectrometry strategy for breast cancer biomarker analysis. Label-free LC-MS/MS was used to identify N-glycosylated proteins in breast tumors, stratified for ER status, from 80 patients who developed or did not develop distant recurrence (DR). Proteins from the LC-MS/MS data were selected to create a 92plex panel for further targeted verification with label-free LC-SRM. All samples were re-run with the targeted 92-plex SRM, and 10 peptides from 10 N-glycosylated proteins were found to be consistently significantly differentially regulated. These proteins may be important for the tumor to obtain the ability to metastasize and for the use as biomarkers. To further validate the clinical accuracy of the 92-plex LC-SRM, we compared our measured levels of HER2 with the clinical evaluation, which showed almost perfect correlation. Finally, the 10 peptides were analyzed at the gene expression level in separate cohort, using publicly available data sets. Five of nine peptides with probes of sufficient quality were validated as prognostic factors, showing the same regulation as the protein level, and in the same subgroups when stratified for ER status.

Experimental procedures Patients and tumors We selected 80 patients from two previous randomized clinical trials37,38, diagnosed with stage II breast carcinoma for which frozen tumors were spared. Tumors were obtained during the primary surgery and before any treatment. Distant recurrence (DR) was developed in 41 patients and 39 were distant recurrence free at least 6.4 years (median 9.4, range 6.4-13.2) after primary breast cancer surgery. Patients were stratified for ER status, as previously assessed by enzyme


5


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 45

immuno assay, iso-electric focusing or immunohistochemistry. Patient characteristics are described in detail in the original studies.37,38 In brief, premenopausal and postmenopausal patients with age 33-77 (mean 58), operated for primary breast cancer between 1983 and 1991, were included. All patients were treated surgically with modified radical mastectomy or breast conserving surgery (BCS) with dissection of the axilla. All patients with BCS were treated with local or locoregional radiotherapy if the disease was spread to regional lymph nodes. All patients received 2 years of adjuvant tamoxifen and mainly no other adjuvant systemic therapy. Tumor samples were directly put on ice during surgery and stored in -80C until analysis. Tumor characteristics were blinded during the experiments. The study was approved by the ethics committee of Lund University (LU 240-01).

HER2 Clinical HER2 over expression of the primary tumor was analyzed for 28 patients on stored tissue blocks at the Pathology department in later retrospective studies. This was not standard care through-out the length of the study, as patients were treated before anti HER2 therapy existed. This explains the high rate of patients with missing clinical HER2 status. HER2 evaluation

was

done

according

to

standard

procedure.

In

brief,

samples

were

immunohistochemically stained and annotated with four levels, 0, 1+, 2+ or 3+. Level 0/1+ patients were considered negative, level 3+ patients were considered positive and for level 2+ patients a fluorescent in situ hybridization (FISH) was made to assess amplification. Patients were considered HER2 positive if amplification was detected. HER2 abundance was also assessed by LC-SRM for two peptides mapped to the HER2 protein: Peptide 1 has the sequence


6

Page 7 of 45

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


NH2-GHBWGPGPTQBVDBSQFLR-COOH

and

peptide

2

has

the

sequence

NH2-

NNQLALTLIDTDR-COOH.

Materials and reagents Sartorious Dismembrator S with a tungsten carbide bowl in a PTFE container (all from Sartorious AG, Tagelswangen, Switzerland) were used for grinding fresh frozen tissue samples. RapiGest and SepPak columns were purchased from Waters (Baden-Daettwil, Switzerland), BCA protein assay and sodium meta-periodate from Thermo Fisher formerly Pierce (Lausanne, Switzerland), Sequencing grade modified trypsin from Promega (Duebendorf, Switzerland), Affi-Prep Hydrazde Hz resin suspension from Bio-Rad (Cressier, Switzerland), Mobicol filter tubes from MoBiTec (Goettingen, Germany), recombinant PNGase F from Roche Applied Sciences (Switzerland), glycerol free PNGase F purified from Flavobacterium meningosepticum from New England Biolabs (through Bio-Conept, Allschwil, Switzerland) and MacroSpin C18 column from Harvard Apparatus (through The Nest Group Inc., Southborough, MA, USA) . For LC-MS, the 5 synthetic retention time reference peptides (NH2-AAVYHHFISDGVRCOOH,

NH2-HIQNIDIQHLAGK-COOH,

NH2-TEVSSNHVLIYLDK-COOH,

NH2-

GGQEHFAHLLILR-COOH, NH2-ITPNLAEFAFSLYR-COOH) were purchased from JPT Peptide Technologies GmbH (Berlin, Germany), Glu-1-Fibrinopeptide B from Sigma (Buchs, Switzerland), coated fused silica capillary from BGB Analytik AG (Boeckten, Switzerland), Magic C18 AQ resin (3 µm-size, 300 A pore size) from Michrom Bioresources Inc. (USA), Biognosys iRT-kit from Biognosys (Schlieren, Switzerland). All other chemicals and reagents were from Sigma (Buchs, Switzerland).


7


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 45

Enriching for N-glycosylated Fresh frozen tissue samples (median: 53 mg, range: 39 mg – 78 mg, interquartile range: 48mg59mg) were put in liquid nitrogen for a few seconds followed by grinding in a PTFE container with a tungsten carbide bowl using the Sartorius Dismembrator S at 3’000 rpm for 30 seconds twice. The container was cooled in liquid nitrogen between the grinding steps. Tissue powder was suspended in 50% (v/v) trifluoroethanol, 50% (v/v) PBS and RapiGest 0.1% (v/v), and the lysate was grinded once more as describe above, transferred to a tube and incubated at 60°C for two hours. Proteins were reduced with final 5 mM DTT at 60°C for 30 minutes and the free cysteine residues were alkylated with final 25 mM iodoacetamide in the dark at room temperature for another 30 minutes. Excess unreacted iodoacetamide was quenched with final 30 mM N-acetyl-cysteine in the dark at room temperature for 15 minutes. Total protein amount of each lysate was determined using the BCA assay prior to dilution of the lysate with 0.1 M ammonium bicarbonate to a final trifluoroethanol concentration of 15% (v/v). Lysate was digested overnight at 37°C with 40 µg Trypsin, digestion completeness was evaluated by SDSPAGE. Digested lysate was stored at -80°C upon further processing. 50% of the total protein amount of the digest was used for the enrichment of N-glycosylated peptides as previously described39, with the difference of using a mix of two PNGaseF reagents. Samples were randomized and were prepared in four sequential batches. For releasing N-glycopeptides from the resin, 5 µl Roche PNGaseF and 0.5 µl NewEngland Biolabs PNGase F were added to the suspended resins prior overnight incubation at 37C. Combined and concentrated eluates were cleaned-up by MacroSpin C18 spin columns (110 x g for 1 minute) using 5% (v/v) acetonitrile 0.1% (v/v) formic acid in water as equilibration and wash solvent and 50% (v/v) acetonitrile


8

Page 9 of 45

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


0.1% (v/v) formic acid in water as elution buffer. The final eluate was concentrated using a vacuum concentrator prior re-suspension for LC-MS analysis.

LC-MS/MS analysis: Data acquisition Samples were dissolved with 400 µl 5% (v/v) acetonitrile, 0.1% (v/v) formic acid and spiked with 5 synthetic retention time reference peptides. 2 µl per sample was analyzed on a Thermo Fisher LTQ-FT (Riehen, Switzerland) three times sequentially, followed by an injection of 80% (v/v) acetonitrile in water injection and a Glu-1-Fibrinopeptide B injection. The injected sample was loaded directly to the analytical column, a 10 cm x 75 µm ID fused silica capillary packed with Michrom Magic C18 AQ 3 particles, and separated with a 60-minute gradient and a flow of 300 nl/min on a Eksigent NanoLC-1D plus (ABSciex GmbH, Brugg, Switzerland) from 4% to 30% solvent B (solvent A: 3% (v/v) acetonitrile, 0.1 % (v/v) formic acid in water / solvent B: 3% (v/v) water, 0.1% (v/v) formic acid in acetonitrile). The gradient was followed by a 10-minute wash segment at 95% solvent B and a 10-minute equilibration segment at 4% solvent B. The column was equilibrated by 30 µl of 4% solvent B prior injection. The LTQ-FT instrument was operated in data-depended selection of precursor ions for CID fragmentation. A precursor scan in the FT mass analyzer for 400-1,600 m/z with a set resolution of 100,000 was used to select the top-3 abundant precursor ions for CID fragmentation in the LTQ with a normalized collision energy of 35. Unassigned and singly charged ion features were rejected for CID. Dynamic exclusion was set to 30 seconds.

LC-MS/MS analysis: Identification of peptides and proteins


9


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 45

CID spectra were searched as fully tryptic or semi-tryptic peptides using Sequest-Sorcerer (SageN, Milpitas, CA, USA) against human IPI database (version 3.38 with 70,757 entries, documentation

is

available

ftp://ftp.ebi.ac.uk/pub/databases/IPI/last_release/current/README),

using a 0.025

at Dalton

precursor mass tolerance. Cysteine residues were statically modified by carbamidomethylation (+57.0215 Dalton), methionines were considered optionally oxidized (+15.9959 Dalton), and asparagines optionally deamidated (-0.9840 Dalton). Search results were validated using the integrated trans-proteomic pipeline of the Sequest-Sorcerer platform using the default options (PeptideProphet followed by ProteinProphet). Identified fully-tryptic peptides containing no methionine and of length between 5 and 25 amino acids, having a deamidated asparagine within the N-glycosylation sequence motif (N-X/P-ST) were used for LC-SRM assay development as described elsewhere.40 Gene ontology analysis as well an over representation test against the human reference list of the identified proteins (p-value ≤ 0.05) was performed using the PANTHER online bioinformatics tool (http://www.pantherdb.org).41,42

LC-MS/MS analysis: Quantification The over 1.5 million acquired MS/MS spectra were further processed for protein quantification using a label-free LC-MS workflow following the 2DDB-framework.43,44 The CID spectra were searched using X!Tandem (version tandem-linux-08-02-01-3/) against a human IPI database (ipi.HUMAN.fasta.20070326) with a monoisotopic mass error of +/- 25 ppm, a maximum charge of 5, carbamidomethylation of cysteines as static modification and optional modification for asparagine deamidation and methionine oxidation. On average 5,250 features were identified in each analysis experiment for a total of 1.25 million features. These features were linked to the


10

Page 11 of 45

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


identified peptide sequences using the MS/MS information, and the sum of all features matched to a protein was used to represent relative protein abundance in each analysis run. 46.4% of the total ion current and 9.4% of all features were annotated using this procedure.

LC-SRM data acquisition and processing LC-SRM was performed on a Thermo Scientific TSQ Vantage instrument (Riehen, Switzerland). Biognosys iRT-peptides were spiked to the each sample.45 The column was equilibrated prior to sample loading with solvent A (solvent A: 3% (v/v) acetonitrile, 0.1% (v/v) formic acid in water / solvent B: 3% (v/v) water, 0.1% (v/v) formic acid in acetonitrile). 1 µl sample was loaded directly on the analytical column (fused silica capillary of 75 µm inner diameter packed 10 cm with Magic C18 AQ from Michrom Bioresources). Peptides were separated using a Thermo Scientific EASY-nLC II system (Riehen, Switzerland) using a flow of 300 nl/min from 5% to 95% solvent B. The gradient was followed by a 8-minute washing step at 100% solvent B. The TSQ Vantage mass spectrometer was operated using the SRM experiment type with a Q1 Peak Width of 0.7 (FWHM) and a constant cycle time of 2.5 seconds. LC-SRM assays for the targeted proteins were developed when creating the SRMAtlas repository, using the LC-MS/MS data presented here and from other sources. Assay development is described in detail in the original SRMAtlas publication.40 In an initial measurement series, all available assays for the targeted proteins were measured on pools of samples in order to select the best detectable peptide per targeted protein. Retention times of the previously developed LC-SRM assays40 were recalibrated to the current instrumental set-up by measuring the Biognosys iRT-peptides in nonscheduled LC-SRM. Candidate LC-SRM assays were recalibrated using the online available iRT-calculator (http://www.biognosys.ch/hrm-mrm-tools/irt-calculator.html) and measured by


11


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 45

scheduled LC-SRM using a 4-minute retention time window around the re-calibrated retention time. Order of sample injection was randomized and each sample was analyzed once. In order to include as many patient samples as possible, we prioritized including more biological replicates than the repeat injection of samples, since the major variance contribution is expected to come from the samples and not from the LC-SRM experiments. RAW files were converted to the open mzXML format using ReadW (version 4.3.1) [http://tools.proteomecenter.org/wiki/index.php?title=Software:ReAdW] and processed by mProphet.46 Briefly, mProphet uses the difference in score distributions between target and explicit decoy data to model an error-rate using machine learning algorithms. A 1% falsediscovery rate was applied for filtering the data points. The Wilcoxon rank-sum test was used on non-normalized peak intensities to assess regulation between groups, with a p-value 1% should be interpreted as peptides with low signal to noise or not present in the sample. However, as we herein aimed to establish a high through-put strategy, relying on LC-SRM for verification, a clear and automated cut-off was required and we chose to use the very stringent filtering of FDR

A Combined Shotgun and Targeted Mass ... - ACS Publications

Recommend Documents