Standardization of a Sample Preparation and Analytical Workflow for

TECHNICAL NOTE pubs.acs.org/jpr

Standardization of a Sample Preparation and Analytical Workflow for Proteomics of Archival Endometrial Cancer Tissue Addie Alkhas,†,‡ Brian L. Hood,‡ Kate Oliver,†,‡ Pang-ning Teng,‡ Julie Oliver,‡ David Mitchell,‡ Chad A. Hamilton,†,‡ G. Larry Maxwell,‡,§ and Thomas P. Conrads*,‡ †

Gynecologic Oncology Service, Department of Obstetrics and Gynecology, Walter Reed National Military Medical Center, Bethesda, Maryland, United States ‡ Women’s Health Integrated Research Center at Inova Health System, Gynecologic Cancer Center of Excellence, Department of Defense, Annandale, Virgina, United States § Department of Obstetrics and Gynecology, Inova Fairfax Hospital, Fairfax, Virgina, United States

bS Supporting Information ABSTRACT: The goal of the present study was to establish a standard operating procedure for mass spectrometry (MS)-based proteomic analysis of laser microdissected (LMD) formalin-fixed, paraffin-embedded (FFPE) uterine tissue. High resolution bioimage analysis of a large endometrial cancer tissue microarray immunostained for the breast cancer type 1 susceptibility protein enabled precise counting of cells to establish that there is an average of 600 cells/nL of endometrial cancer tissue. We sought to characterize the peptide recovery from various volumes of tissue gathered by LMD and processed/digested using the present methodology. We observed a nearly linear increase in peptide recovery amount with increasing tissue volume dissected. There was little discernible difference in the peptide recovery from stromal versus malignant epithelium, and there was no apparent difference in the day-to-day recovery. This methodology reproducibly results in 100 ng of digested peptides per nL of endometrial tissue, or ∼25 pg peptides/endometrial cancer cell. Results from liquid chromatography (LC) MS/MS experiments to assess the impact of total peptide load on column on the total number of peptides and proteins identified from FFPE tissue digests prepared with the present methodology indicate a demonstrable increase in the total number of peptides identified up to 1000 ng, beyond which diminishing returns were observed. Furthermore, we observed no impact on the peptide identification rates from analyses of equivalent peptide amounts derived from lower volume LMD samples. These results show that this single-tube collection-to-injection proteomics (CTIP) workflow represents a straightforward, scalable, and highly reliable methodology for sample preparation to enable high throughput LMD-MS analysis of tissues derived from biopsy or surgery. KEYWORDS: formalin-fixed, paraffin-embedded, tissue, laser microdissection, proteomics, endometrial cancer

’ INTRODUCTION The growth and invasion of cancerous cells into the stroma of a tissue is a complex and poorly enumerated event at the molecular level. Increasing evidence suggests that a great deal of loco-regional biochemical transactions in the form of differential protein expression and signaling occur that either support or discourage tumor growth, invasion, metastasis, angiogenesis and metastatic seeding.1 7 This “proteomic” communication is thought to take the form of intercellular adhesions and signaling and originates not just from tumor cells as mounting evidence clearly supports the notion that the protein complement from stromal cells and the ECM influences the carcinogenic and metastatic potential of a variety of tumor types. The concept of the influence of the stroma in the tumor microenvironment is not a new one. Indeed, it is the basis of the seed and soil concept. Quite amazingly, in 1889 Dr. Stephen Paget published an article describing his thoughts and observations where he noted that “The best work in the pathology of r 2011 American Chemical Society

cancer is now done by those who are studying the nature of the seed. They are like scientific botanists; and he who turns over the records of cases of cancer is only a ploughman, but his observation of the properties of the soil may also be useful.”8 A growing appreciation of the contribution of the stroma, along with the ECM components, to tumor initiation, progression and metastasis has been a subject of increasing focus in the past decade through detailed molecular studies in specific tumor types. It is known that the stromal microenvironment in many human tumors is quite different from the corresponding stroma of “normal” tissue.9 Studies of human cancers of the breast, pancreas, colon, prostate and lung have demonstrated reactive changes in the stroma that lend to a permissive and supportive environment for tumor growth.10 The changes observed are not limited to morphology but include accumulation of connective Received: August 11, 2011 Published: September 19, 2011 5264

dx.doi.org/10.1021/pr2007736 | J. Proteome Res. 2011, 10, 5264–5271

Journal of Proteome Research tissue, proliferation of fibroblasts and changes in gene expression that are influenced by biochemical cross-talk between stroma and tumor cells. Highlighting this influence, a recent report reveals that changes in the expression of certain genes in the stroma actually coincide with changes in expression of the same gene in tumor cells as a function of progression.11 This investigation demonstrated a remarkable down-regulation of the chloride intracellular channel isoform 4 (CLIC4) in the epithelium of multiple tumor types with a corresponding up-regulation of this protein in the stroma that correlates with the state of tumor progression. Differential proteomic analysis of representative stromal and cancerous tissue by mass spectrometry (MS) is likely to prove useful for detailing the complexities associated with this complex microenvironment. The information obtained can help construe the function of individual proteins and establish the composition of functional units, protein protein interactions, and dysregulated protein pathways and networks. Proteomics promises to revolutionize the ability to provide detailed molecular classification of tumor tissue and, ultimately, enable personalized assessment to guide molecularly targeted therapeutic regimens. Although proteomic investigations are likely to offer considerable benefits to the clinical management of cancer, there are substantial challenges related to standardization of tumor specimen preparation for downstream analysis. These challenges are particularly important to overcome considering the large amount of resources (e.g., personnel and instrument time, tissue specimens, reagents, etc.) that are leveraged in conducting biomarker discovery projects. We sought to standardize a number of parameters that are poorly defined in the sample preparation and processing pipelines for MS-based proteomic analysis of laser microdissected (LMD) tumor tissue, a technique that offers considerable promise to enable detailed investigations of the epithelial and stromal compartments that comprise the tumor microenvironment. In an extension of previous techniques developed in our lab and others,12 22 we characterized critical sample preparation parameters for conducting LMD-MS-based proteomics utilizing formalin-fixed, paraffin-embedded (FFPE) tissue blocks from surgically resected uterine specimens. These experiments describe important parameters such as the relationship between the tissue input amount and the peptide identification rate, the sensitivity of the LC MS/MS result to the trypsin to tissue ratio for digestion, the correlation between the volume of tissue collected by LMD and the cellular content, and the correlation between this volume of tissue and the protein/ peptide yield. Our refined approach provides critical metrics toward qualifying and quantifying key sample preparation checkpoints to support a single vessel collection-to-injection proteomic (CTIP) workflow for analysis of tissues derived from biopsy or surgery.

’ MATERIALS AND METHODS Tissue Specimens

Endometrial cancer FFPE tissue specimens were obtained from the Armed Forces Institute of Pathology under an IRB approved research protocol. Tissue sections (8 μm thick) were cut by microtome and placed on PALM MembraneSlides (Carl Zeiss, Inc., Thornwood, NY) for LMD. All membrane slides were prepared by cross-linking for 30 min using a UV cross-linker (Spectronics Corporation, Westbury, NY).

TECHNICAL NOTE

Figure 1. (A) High resolution scanned image of an endometrial cancer tissue microarray (TMA) slide following markup of the annotations for each core as visualized within the Aperio ImageScope platform (magnification 1.1). (B) Representative single core markup from the Aperio ImageScope Nuclear-BRCA1 algorithm following automated bioimage analysis of the endometrial cancer TMA (magnification 20). Each identified nucleus was designated by staining intensity using a standard 0 3+ scale employed in immunohistochemistry, and visually depicted using a color scale (blue = 0, yellow = 1+, orange = 2+, red = 3+) in the markup overlay.

Calculation of Cell Population per Volume of Tissue

Bioimage analysis of an endometrial cancer tissue microarray (TMA) was used to estimate the cellular composition of the endometrial cancers evaluated in this study. The TMA was constructed from 0.7 mm cores obtained from banked endometrial tumors from 155 patients representing a range of histologies, stages and grades of endometrial cancer. In this case, the TMA slides were stained (from another unrelated study) using the mouse monoclonal antibody against the breast cancer type 1 susceptibility protein (anti-BRCA1, Catalog # OP92, EMD Chemicals, Gibbstown, NJ), validated for use in immunohistochemical-paraffin (IHC-P) applications. The TMA slides were digitized to 20 magnification using an Aperio Digital Pathology Services ScanScope XT (Dublin, Ireland). For each core, between 3 and 10 annotations of the cancer areas were delineated by a single contributor and independently reviewed by a boardcertified pathologist for accuracy (Figure 1A). The annotations were selected to maximally exclude stroma and other noncancerous tissue. Control cores, missing cores, spacers and cores without acceptable cancer tissue were designated with a single, empty annotation for place holding. 5265

dx.doi.org/10.1021/pr2007736 |J. Proteome Res. 2011, 10, 5264–5271

Journal of Proteome Research

TECHNICAL NOTE

Table 1. Peptide and Protein Identification Results for Analysis of 63 nL of Laser Microdissected Tissue Digested with Varying Amounts of Trypsin Trypsin to Protein Ratio 1:10 p1651 p1654

p1657

a

4325

3432

4080

Total Peptides, 2b Total Peptides, 3c

4351 4424

3650 3525

4210 4105

Average

4367

3536

4132

Relative Standard Deviation

1.2

3.1

1.7

Total Proteins

1531

1367

1592

Proteins g2 Peptides

778

667

768

specimen Total Peptides, 1

Figure 2. Total peptide recovery from defined volumes of laser microdissected (LMD) formalin-fixed, paraffin embedded tissue. Defined volumes of stromal (S) and cancerous (Ca) epithelial tissue were collected by LMD (in duplicate) processed, digested and assayed by the bicinchoninic acid assay. A biological replicate representing two cancerous epithelial samples of 63 nL of tissue collected by LMD was performed on a separate day to evaluate day-to-day reproducibility (CA 3 2). Error bars represent the standard deviation observed from three independent peptide concentration measurements.

Automated bioimage analysis was performed on the annotated regions using the Aperio Digital Pathology Services Spectrum information management system in conjunction with the ImageScope (version 11.0.2.725) Nuclear-BRCA1 algorithm. Data captured from the bioimage analysis included total area of analysis for each annotation, average nuclear size within a specific annotation, and total identified nuclei for the annotation. Additionally, each identified nucleus was designated by staining intensity using the typical 0 3+ scale employed in immunohistochemistry, and visually depicted using a color scale (blue = 0, yellow = 1+, orange = 2+, red = 3+) in the markup overlay. A representative postbioimage analysis core markup is shown in Figure 1B. The results of the automated bioimage analysis were exported from the Aperio Spectrum information management system into Mathworks MATLAB (version 7.12.0.635; R2011a) for bioinformatics.

Trypsin to Protein Ratio 1:20 p1652 p1655

p1658

Total Peptides, 1

4143

3414

3795

Total Peptides, 2

4296

3408

3815

Total Peptides, 3

4249

3411

3875

Average

4229

3411

3828


1.9

0.1

1.1

Total Proteins Proteins g2 Peptides

1474 750

1335 641

1503 734

Trypsin to Protein Ratio 1:50 p1653 p1656

p1659

Total Peptides, 1

4092

3461

3749

Total Peptides, 2

4193

3434

3828

Total Peptides, 3 Average

4170 4152

3450 3448

3895 3824


1.3

0.4

1.9

Total Proteins

1500

1324

1507


735

651

730

specimen

specimen

a

First LC MS/MS analysis. b Second LC MS/MS analysis. c Third LC MS/MS analysis.

spectrophotometry (NanoDrop 2000, ThermoScientific) using the bicinchoninic acid assay (BCA, Pierce, Rockford, IL).

Laser Microdissection

Prior to LMD, all membrane slides were deparaffinized in xylene and rehydrated through graded ethanol washes and stained in Mayer’s hematoxylin. After a minimum of 24 h of air drying, LMD was performed (LMD 6500, Leica Microsystems, Wetzlar, Germany) to capture defined regions from the tissue microenvironment, including glandular structures as well as representative stromal tissue. Tissue was collected into 45 μL of distilled water and stored at 80 °C. Digestion of Formalin-Fixed, Paraffin-Embedded Tissue

Tissue was allowed to thaw at ambient temperature for approximately 15 min. Samples were made to a final concentration of 100 mM ammonium bicarbonate (AmBic)/20% acetonitrile (MeCN). The samples were incubated at 95 °C for 1 h, followed by 2 h at 65 °C after which sequencing grade porcine trypsin was added at 50 ng per 63 nL of tissue unless indicated otherwise. Samples were incubated for 16 h at 37 °C, after which they were dried by vacuum centrifugation. Tissue digests were resuspended in 30 μL of 25 mM AmBic, from which 5 μL was used to quantify total peptide concentration by

Liquid Chromatography Tandem Mass Spectrometry

Tissue digests were analyzed in duplicate for the peptide recovery experiment and in triplicate for all other experiments. All LC MS/MS analyses were performed on a nanoflow LC system (Easy-nLC, ThermoFisher Scientific, San Jose, CA) coupled online with a LTQ-Orbitrap Velos MS (ThermoFisher Scientific). All of the samples were resolved on a 100 μm I.D. 360 μm O.D. 20 cm long capillary column (Polymicro Technologies, Phoenix, AZ), which was slurry packed in house with 5 μm, 300 Å pore size C-18 silica-bonded stationary phase (Jupiter, Phenomenex, Torrance, CA). Following precolumn and analytical column equilibration, each sample was loaded onto a 2 cm reversed-phase (C-18) precolumn (ThermoFisher Scientific) at 2 μL/min for 6 min with mobile phase A (0.1% formic acid in water). Peptides were eluted at a constant flow rate of 200 nL/min by development a linear gradient of 0.33% mobile phase B (0.1% formic acid in acetonitrile) per min for 120 min and then to 95% B for an additional 15 min. The column was washed for 15 min at 95% B and then quickly brought to 100% A for the next sample injection. The LTQ-Orbitrap Velos MS was 5266



TECHNICAL NOTE

Table 2. Peptide and Protein Identification Results from LC MS/MS Analysis of 500 ng Peptide from 31.5 nL of Laser Microdissected Endometrial Cancer Tissue 31.5 nL tissue input specimen

p1639

p1640

a

3809

3562

Total Peptides, 2b

3866

3677

Total Peptides, 3c

4060

3666

Average

3912

3635

Relative Standard Deviation Total Proteins

3.3 1528

1.7 1471


750

693

Total Peptides, 1

a

First LC MS/MS analysis. b Second LC MS/MS analysis. c Third LC MS/MS analysis.

Figure 3. Metrics of peptide and protein identifications from liquid chromatography-tandem mass spectrometry at varying amounts of total peptide analyzed.

configured to collect high resolution (R = 60 000 at m/z 400) broadband mass spectra (m/z 375 1800) using the lock mass feature for the polydimethylcyclosiloxane (PCM) ion generated in the electrospray process (m/z 445.12002). Mass spectrometric conditions were set as follows: electrospray voltage, 1.9 kV; no sheath and auxiliary gas flow; capillary temperature, 220 °C; S-Lens RF level, 69%. The ion selection threshold for the Orbitrap (MS) was set at 1e6 with a maximum ion accumulation time of 500 ms. The twenty most abundant ions were selected for MS/MS in the high-pressure linear ion trap (MS/MS) with the following settings: ion threshold, 5000; minimum intensity, 3000; maximum ion accumulation time, 25 ms; activation time, 10 ms. Dynamic exclusion (60 s) was utilized to minimize redundant selection of peptides for MS/MS. Bioinformatic Analysis

Tandem mass spectra were searched against the UniProt human protein database (01/2011) from the European Bioinformatics Institute (http://www.ebi.ac.uk/integr8/) using Mascot Daemon (Matrix Science Inc., Boston, MA). The data were searched with a precursor mass tolerance of 10 ppm and a fragment ion tolerance of 0.6 Da. Methionine oxidation (15.99492) was set as a dynamic modification and a maximum of two missed cleavages were allowed. An automatic decoy search was performed on all raw files and peptides were filtered using an ion score cutoff of 30 resulting in a false peptide discovery rate of approximately 2%. In cases where peptides were identified in more than one protein sequence in the database, protein identifications were compiled based on the first UniProt Accession number listed in the protein match results for each peptide.

’ RESULTS AND DISCUSSION Our goal was to use design of experiments (DOE) to determine and establish parameters toward a best practice standard operating procedure for the analysis of laser microdissected cells from FFPE tissue utilizing a single-vessel CTIP workflow. The first parameter established was an estimate of the

Table 3. Peptide and Protein Identification Results from LC MS/MS Analysis of 500 ng Peptide from 63 nL of Laser Microdissected Endometrial Cancer Tissue 63 nL tissue input day processed

day 1

day 2

specimen

p1636

p1637

p1647

p1648

Total Peptides, 1a

4633

4327

4443

3954

Total Peptides, 2b

4732

4290

4372

4039

Total Peptides, 3c

4805

4125

4117

4001

Average

4723

4247

4310

3998


1.8

2.5

4.0

1.1

Total Proteins Proteins g2 Peptides

1720 836

1580 783

1599 799

1544 735

a First LC MS/MS analysis. b Second LC MS/MS analysis. c Third LC MS/MS analysis.

number of cells contained per unit area of thin tissue section used for LMD. To this end, we utilized an endometrial cancer TMA that was constructed from 155 patient tissue specimens and stained for the BRCA1 protein (Figure 1), which is localized to the nucleus and hence representative of a cellular unit. An average of 126 tissue cores were accepted for analysis over the three TMA slides utilized. For each of the slides and overall, the nuclear area comprised approximately 33% of the total analysis area over all of the TMA core annotations. A total of 176 263 nuclei were counted over a total area of 25 mm2 for a nuclear density of 7051 nuclei/mm2. Assuming cells are spheroid and using the formula for the volume of a sphere (4/3πR3), the diameter for an individual endometrial cancer cell was calculated to be approximately 13 μm, which is within biological plausibility. In our preparatory workflow, 8 μm thick tissue sections are typically used for LMD and therefore the tissue thin section is comprised of no more than an average of 0.6 nuclei per unit thickness in volume. Based on these determinations, we estimate that approximately 600 tumor cells are collected per nL of tissue harvested by LMD. Since our sample preparation methodology relies on the digestion of formalin-fixed proteins directly from tissue, we investigated whether the amount of trypsin used would have an impact on the overall recovery of peptides from digestion. We 5267



TECHNICAL NOTE

Figure 4. Plot of the gene ontology distribution of proteins identified from laser microdissected (LMD) formalin-fixed, paraffin-embedded (FFPE) endometrial cancer cells classified by cellular component. Three technical replicate LC MS/MS analyses of 500 ng total peptide digest were conducted from cells collected by LMD from four FFPE endometrial cancer specimens. Samples p1636 and p1637 represent 63 nL of malignant epithelium collected by LMD as the input for sample preparation and digestion, whereas 31.5 nL was harvested from samples p1639 and p1640. In addition, all accession entries for the human proteome (UniProt human protein database [01/2011] from the European Bioinformatics Institute [http://www.ebi.ac.uk/integr8/]) are plotted for comparison.

therefore assessed the optimal amount of trypsin for efficient digestion of proteins in LMD FFPE samples. Sixty-three nanoliters of tissue was collected by LMD from nine different malignant endometrial FFPE specimens. Estimating 7.5 μg total protein from 63 nL of tissue (see Figure 2 and below), three different trypsin to total protein ratios (1:10, 1:20 and 1:50) were used to study the impact on peptide (and protein) identification metrics by LC MS/MS. Three technical replicates of each of the three tissue samples digested at each of the three ratios of trypsin to total protein (a total of 9 tissue samples) were analyzed by LC MS/MS. The results (Table 1, Supplemental Table 1, Supporting Information) indicate similar peptide and protein identification rates from digestion at trypsin to protein ratios above that commonly utilized in the majority of proteomics experiments (e.g., a ratio of 1:50). Not surprisingly, the LC MS basepeak chromatograms of samples digested with trypsin to protein ratios above 1:50 have a number of intense and broadly eluting peaks that are evident throughout the gradient that arise from trypsin peptides resultant from autolysis (data not shown). We therefore used a ratio of trypsin to protein of 1:50 for all subsequent experiments. We sought to characterize the peptide recovery from various volumes of tissue gathered by LMD and processed/digested using our present method. Three samples each representing 7, 31.5, and 63 nL were obtained by LMD from malignant epithelium, along with 7 nL of hyperplastic endometrial stroma

to provide an additional comparison of recovery from these two tissue compartments. As expected, peptide amounts recovered increased nearly linearly with volume of tissue dissected (Figure 2). Furthermore, there was no difference in the peptide amounts recovered from stromal (S) versus malignant epithelium (Ca 1) and there was no difference in the day-to-day recovery (Ca 3 1 vs Ca 3 2). These data demonstrate that this methodology reproducibly results in approximately 100 ng of digested peptide recovered per nL of endometrial tissue, or ∼25 pg peptides/endometrial cancer cell. By contrast, in a recent publication by Wisniewski et al. that employed a filter-aided sample preparation (FASP) workflow, 1 μg of total peptides were obtained from 30 nL of microdissected tissue, which represents a 3-fold lower recovery than our current method.23 In a response to the Wisniewski paper regarding their spin filter-based sample preparation, Liebler and Ham24 compared the spin filter method to methods involving digestion of proteins from short SDSPAGE, as well as a single tube trifluoroethanol (TFE) extraction procedure,25 and noted that for high protein loads of 50 μg and greater, peptide and protein identifications were comparable, but at low protein loads the FASP method yielded only 31% of the identifications as found with the single tube TFE method. A number of LC MS/MS analyses were conducted to assess the impact of total peptide load on column on the total number of peptides and proteins identified from FFPE tissue digests prepared with this methodology. Two samples each representing 5268


Journal of Proteome Research 63 nL of tissue were collected from the same malignant endometrial tissue. Two technical replicate LC MS/MS analyses were conducted on aliquots ranging from 100 to 1000 ng total peptide. There was a demonstrable increase in the total number of peptides identified up to 1000 ng analyzed (Figure 3) where 9327 total peptide identifications resulted from the two technical replicates (Supplemental Table 2, Supporting Information), beyond which diminishing returns were observed based on an analysis of 2000 ng (data not shown). Results from LC MS/MS of equivalent peptide amounts derived from different total volumes of tissue were also compared to assess whether the amount of LMD tissue impacts the aggregate peptide and protein identification metrics. In this set of experiments two samples each representing 31.5 and 63 nL of malignant epithelium from endometrial cancer tissue were collected by LMD. Each of the two samples from the two volumes of endometrial cancer tissue were digested according to the detailed methodology presented herein and three technical LC MS/MS replicate analyses were performed, each of which consumed 500 ng of total peptide digest. An average of 3774 total peptides and 1500 proteins were identified per LC MS/MS analysis (Table 2, Supplemental Table 3) of 500 ng of total peptides derived from 31.5 nL of malignant endometrium. By contrast, an average of 4320 and 1611 total peptides and proteins, respectively, were identified per LC MS/MS analysis (Table 3, Supplemental Table 3, Supporting Information) of 500 ng of total peptides derived from 63 nL of malignant endometrium. These results indicate an overall low variance in peptide (

Standardization of a Sample Preparation and Analytical Workflow for

Recommend Documents