Protein profiling of formalin fixed paraffin embedded tissue: Identification of potential biomarkers for pediatric brainstem glioma (original) (raw)

. Author manuscript; available in PMC: 2014 Jan 30.

Published in final edited form as: Proteomics Clin Appl. 2008 Jun;2(6):915–924. doi: 10.1002/prca.200780061

Abstract

Little is known about the molecular characteristics of pediatric brainstem gliomas (BSG), which continue to have a dismal prognosis. Targeted molecular strategies are limited due to rarity of biopsy BSG specimen coupled with obstacles associated with the analyses of formalin-fixed paraffin- embedded (FFPE) autopsies. The objective of this study was to develop methodologies to successfully identify the proteome profile from these archived FFPE specimens. Peptides were extracted from both tumor and adjacent normal FFPE brainstem specimen and quantified using 18O proteolytic labeling strategy and LC-MS/MS analysis. The ingenuity pathway analysis software was used to elucidate interactions amongst differentially expressed proteins. We identified 188 proteins of which 54 (29%) were found up-regulated (≥1.5-fold) in BSG compared to normal sections. Of these, 15 (28%) proteins have previously been reported as potential biomarkers for supratentorial malignant gliomas, while the rest appear to be exclusive to pediatric BSG. Because the majority of differentially expressed proteins are unique to BSG, we conclude that pediatric BSG is distinct from supratentorial gliomas. To the best of our knowledge, this is the first proteome profile of pediatric BSG, which may facilitate discovery of novel therapeutic targets for early diagnostics and improving prognostics.

Keywords: Formalin fixed paraffin embedded, Ingenuity pathway analysis, Pediatric brainstem

1 Introduction

Pediatric brainstem glioma (BSG) is the most difficult childhood cancer to treat, largely due to the inoperability and diffuse nature of the lesion. Apart from conventional radiation therapy, which typically results in only transient symptomatic relief, other treatments have failed to improve outcome, including high doses (up to 78 Gy) of radiation [1] and the addition of chemotherapy that has been effective for other types of brain tumors [2–4]. In fact, for diffused gliomas, despite numerous clinical research trials over the last three decades investigating the use of experimental chemotherapy, radiation and biologic therapy, the overall survival rate of approximately 10% has remained unchanged [5, 6].

BSG predominantly occurs in children between the ages of six to nine years and accounts for 10–20% of all pediatric CNS tumors [7, 8]. The pons is the most common brainstem location of BSG; however, it can also arise in the cervicomedullary junction, midbrain, or the tectum. BSG are categorized into two main groups: diffusely infiltrative and focal [9, 10]. Diffusely infiltrative BSG spreads throughout the pons with occasional dissemination to other parts of the brainstem. Focal brainstem gliomas are, however, localized tumors typically occurring in the midbrain or medulla.

Recent advances in the treatment of other refractory cancers have been accomplished through drug targeting of identified biomarkers that are specific to the type of tumor. To date, however, there are no known biomarkers specific to BSG. Despite advancement of proteomics and genomics applications, little progress has been made towards understanding the molecular constituents of BSG primarily because of the relative lack of biological samples for conducting research studies of this kind. Because this tumor has a typical appearance on magnetic resonance imaging (MRI), and because it is often difficult to safely perform biopsies of BSG, obtaining tumor tissue for histological confirmation of the diagnosis is generally not clinically warranted. Thus, there is a lack of available fresh-frozen tissue for the molecular characterization of these tumors. Furthermore, the complexities associated with protein profiling of formalin-fixed paraffinembedded (FFPE) autopsy specimens have similarly limited performing analyses for biomarker discovery. New methodologies allowing for the identification of BSG-specific biomarkers are therefore greatly needed.

Formalin fixation has been long used to study molecular interactions. Histone-histone and histone-DNA complexes for example, were studied by formation and then reversal of formalin cross-links [11–13]. Cross-linking experiments have also been insightful into the interaction of anticancer drugs and DNA molecule [14, 15]. Formalin-fixed autopsy samples are an excellent resource for genomic and proteomic analyses. During the past decades, methods for reversing cross-links of FFPE samples have improved, specifically for antibody retrieval in immunohistochemistry assays [16–20]. Recent advances in proteomics field have also led to the successful extraction and analysis of peptides from FFPE tissues [21–23]. In an attempt to discover potential BSG markers for therapeutic targeting, as well as define the molecular signature of the tumor, we used FFPE pediatric BSG archival tissues for protein identification and quantification. Our analyses show the feasibility of profiling proteins from archival FFPE brain tissues and obtaining qualitative and quantitative molecular insights into BSG. Here, we have successfully conducted comparative protein profiling of normal versus BSG brainstem area obtained from the same FFPE patient sample and generated a list of 54 differentially expressed proteins. Thus, we present the first proteome profile of BSG, suggesting potential biomarkers and molecular signatures of disease progression for possible therapeutic intervention as well as describe a technique that may be utilized broadly for all FFPE cancer specimens.

2 Materials and methods

2.1 Human subjects

FEPE BSG autopsy samples were provided by the Children’s National Medical Center. Institutional Review Board (IRB) was obtained to perform evaluations on these samples. All identifiers were removed from the samples prior to evaluation.

Two FFPE BSG autopsy samples were analyzed. Hematoxylin and eosin (H&E)-stained FFPE sections (5 µm thick), were examined by a neuropathologist using light microscopy to identify the area of the brainstem involved by tumor versus the adjacent normal areas. All the cases selected were astrocytoma. For each case, the best possible representative area of tumor was highlighted by the neuropathologist with a marker such that the identical area on the unstained slide could be located by overlapping the H&E-stained section with the unstained section. The unstained sections were all 7 µm in thickness and were prepared on Superfrost Plus microscope glass slides (Fisher Scientific, Houston, TX). Ten slides were used to extract tumor and normal areas for each experiment. Slides were heated in a bench-top oven at 65°C for 60 min to remove excess wax. To ensure maximum wax removal, slides were washed for two rounds of xylene (5 min each), two rounds of 100% fresh ethanol (5 min each), two rounds of 95% ethanol (3 min each), two rounds of 90% ethanol (3 min each), two rounds of 75% ethanol (3 min each) and then rehydrated in water for 30 s.

Delineated areas of interest were then removed directly from the slides as explained here. To remove tissue sections, 10–15 µL of extraction buffer (100 mM ammonium bicarbonate and 30% ACN) was directly deposited on marked tumor or normal areas. Hydrated tissue was then scraped off using the pipette tip, suctioned off the slide, and deposited into 1.5- mL polypropylene tubes (Eppendorf, Westbury, NY). The tumor tissue subsequently removed by pipetting resulted in the recovery of the most representative area of tumor for analysis. To reverse cross-links, tubes containing the tissue were placed in a bench-top oven at 95°C for 30 min followed by 2-min incubation on ice and 3 h at 65°C. Samples were immediately placed in vacuum centrifuge and completely dehydrated.

2.3 Proteolytic 18O-labeling and peptide extraction

Differential labeling by stable isotope was performed using 18O proteolytic labeling strategy [24]. Briefly, dehydrated tissue extracts were digested by trypsin either in 18O-enriched (97%) water (Sigma-Aldrich, St. Louis, MO) or in 16O-water. Two micrograms of horse cytochrome C (CYC; Sigma-Aldrich) was added in each sample as an internal standard. For data confirmation and validation, reverse labeling was also performed where normal extracts were digested in the presence of heavy 18O and tumor extracts were digested in the presence of light 16O-water. All digestions were performed overnight in a water bath set at 37°C in the presence of 4.5 µg MS grade trypsin (Promega, Madison, WI). To ensure a complete digestion and incorporation of 18O atoms in the C termini of tryptic peptides, additional trypsin (3.5 µg) was added next day and digestion was continued for an additional 5 h.

2.4 Quality control of peptides extracted from FFPE tissue and of 18O labeling

After digestion, samples were spun in a bench top centrifuge at maximum speed to pellet cellular debris. Supernatant was then removed and stored at −80°C until further processing. Two microliters of supernatants from each sample was desalted using C18 ZipTip micropipette tips (Millipore, Bedford, MA) following the manufacturer’s User Guide. The peptides were eluted from the ZipTip in 10 µL of ACN/0.1% TFA (70:30 v/v). Peptide solution (0.3 µL) was mixed with 0.3 µL of matrix solution (50 mM CHCA in ACN/0.1% TFA 70:30 v/v) and spotted on the MALDI plate. MS analysis was performed on a 4700 ABI TOF-TOF mass spectrometer (Applied Biosystems, Foster City, CA) in a positive reflectron mode. A mixture of standard peptides was used to externally calibrate the instrument. The high resolution of the ABI-4700 MALDITOF- TOF and its sensitivity helped verification of both the quality of extracted peptides and the incorporation rate of 18O atoms at the C termini of each tryptic peptide. Inspected samples were then completely dried using a vacuum centrifuge, dissolved in 4 µL of 0.1% TFA and immediately analyzed by LC-MS/MS as described below.

2.5 Nanoflow LC-MS/MS

Differentially labeled peptides from normal and tumor sections were mixed at a 1:1 ratio (4 µL each) and an aliquot (6 µL) was analyzed by LC-MS/MS using an LC-Packing system (DIONEX Ulti-Mate™ Capillary/Nano LC System, Dionex, Sunnyvale, CA) connected to a Linear IT (LTQ) mass spectrometer (Thermo Electron, San Jose, CA). Each sample was injected via an autosampler and loaded onto a C18 trap column (300 µm×1 mm, LC Packing) for 6 min at a flow rate of 10 µL/min. The sample was subsequently separated by a C18 RP column (75 µm × 15 cm, Vydac, Columbia, MD) at a flow rate of 300 nL/min. The mobile phases consisted of water with 0.1% formic acid (A) and 90% ACN with 0.1% formic acid (B), respectively. A 240-min linear gradient from 5 to 50% B was typically employed. Separated peptides were introduced into the mass spectrometer via a 10-mm silica tip (New Objective, Ringoes, NJ) adapted to a nanoelectrospray source (Thermo Electron). The spray voltage was set at 1.7 kV and the heated capillary at 160°C. The LTQ was operated in data-dependent mode in which one cycle of experiments consisted of one full-MS survey and subsequently three sequential pairs of intercalated zoom scans and MS/ MS experiments. The targeted ion counts in the IT during full-MS, zoom scan, and MS/MS were 30 000, 3000 and, 10 000, respectively. Peptides were fragmented in the trap using CID with the collision gas (helium) pressure set at 1.3 mTorr and the normalized collision energy value set at 35%.

Protein identification was performed with the BioWorks 3.2 software (Thermo Electron, Waltham, MA). Briefly, each file was searched against the human Swiss-Prot database (version 52.0, release date of March 2007), which contained 15 945 protein sequences. Horse CYC FASTA sequence (Swiss-Prot ID P00004) was added manually to the Swiss-Prot library, with the following possible protein modifications: 16-Da shift for oxidized Met and 4-Da Cterminal modification (18O incorporation). The acceptance criteria for peptide identification were set as follows: a DeltaCn (DCn).0.1, a variable threshold of Xcorr versus charge state (Xcorr = 1.9 for z = 1, Xcorr = 2.5 for z = 2, and Xcorr = 3.5 for z =3) and a peptide probability based score with a p value <0.001. False discovery rate (FDR) was determined at the protein level by searching a reversed FASTA format of human Swiss-Prot database using the same filtration criteria. FDR was calculated to be less than 3.2% using the following formula: FDR = [2 × (Reverse Database Protein Hits)/(Reverse + Forward Database Protein Hits)] × 100.

2.7 ZoomQuant analysis

To measure the intensity ratios of labeled and unlabeled peptide pairs, we used the ZoomQuant software developed by Halligan and colleagues [25]. The generated SEQUEST output files were loaded into the Epitomize filter to generate a colon-delimited text file that contained the list of identified proteins, the sequence and atomic composition of the identified peptides with scan number, charge state, Xcorr, and mass data. This file was then loaded into the ZoomQuant software along with a file containing zoom scans that were extracted from the corresponding raw data file using a visual basic RawBitz script. Zoom scan information was then matched to the SEQUEST-identified peptides, after which labeled and unlabeled pairs were detected based on their amino acid composition and theoretical masses. Zoom- Quant software takes into account partial (one 18O molecule) as well as full (two 18O molecules) incorporation of heavy oxygen. Ratios were determined from the peak areas of the isotope clusters of the labeled and unlabeled peptides. Obtained ratios were then normalized to the horse CYC ratios as described here.

2.8 Data normalization and assessment of peptide concentration

Horse CYC was used as the internal control. Two micrograms of purified horse CYC was added to each sample prior to proteolytic assay. Peptide concentration estimation was done using spectra generated by MALDI-TOF for both the normal and tumor regions prior to mixing the samples. Consequent to analyzing samples with MALDI-TOF, generated spectra were searched and the following three peptide peaks of CYC were selected: MIFAGIKK (m/z = 907.8), TGPNLHGLFGR (m/z = 1168.1), and IFVQKCAQCHTVEK (m/z = 1633.3) for 16O-labeled (Fig. 2A, Insets A1, A2, and A3) and 18O-labeled peptides (Fig. 2B, Insets B1, B2, and B3). Relative peptide concentrations were estimated by dividing the total ion count value of all observed peaks by the average ion count of the three CYC peaks. These relative concentrations were then used to determine the normalization factor we used to adjust for the ZoomQuant (tumor to normal) ratios. Average fold change was calculated based on values generated by the two samples as well as reverse labeling experiments.

Figure 2.

Figure 2

Analyses of FFPE extracted peptides using MALDI TOF-MS. Aliquots of 16O- and 18O-labeled peptides extracted from FFPE samples were separately analyzed by MALDI TOF-MS to verify the efficiency of peptide extraction (Panels A and B). Relative amount of extracted peptides in the presence of either 16O (A), or 18O (B) were estimated by dividing the TIC value of all observed peaks by the average ion count of the three CYC peptide peaks: MIFAGIKK (m/z = 907.8, Inset A1), TGPNLHGLFGR (m/z = 1168.1, Inset A2), and IFVQKCAQCHTVEK (m/z = 1633.3, Inset A3). Compared to 16O (Insets A1, A2, and A3) incorporation, labeling with 18O resulted in a shift of molecular masses of the three CYC peptides (B, insets B1, B2, and B3).

3 Results

3.1 Assessment of peptide extraction from (FFPE) tissue and of 18O-labeling efficiency

We have implemented and optimized a method for the extraction of peptides from FFPE BSG samples (Fig. 1). This improved protocol includes three main steps, (i) peptide extraction, (ii) stable isotope labeling using enriched 18O water, and (iii) proteome profiling. Each step of this process was checked for quality control (QC) to ensure the validity of the process.

Figure 1.

Figure 1

Flowchart of the implemented methodology for the extraction and the analyses of peptides from BSG FFPE samples. The overall methodology consists of three major steps: (i) removal of paraffin by heating samples and a series of washes followed by extraction of tryptic peptides (information acquisition); (ii) addition of 2 mg horse CYC to each sample as internal control followed by 18O or 16O proteolytic labeling and control quality by MALDI-TOF analysis; (iii) in the last step, peptides extracted from the tumor and the normal regions are mixed and subjected to LC-MS/ MS analysis.

Before mixing 18O-labeled and unlabeled samples for proteome profiling, we checked both the yield of peptide extraction from FFPE tissue and the efficiency of 18O incorporation during proteolysis (see Section 2). Each sample was spiked with 2 µg horse CYC before proteolysis. Following the completion of proteolysis, 2 µL of tryptic digested peptides from 16O- (Fig. 2A) and 18O-labeled peptides (Fig. 2B) were separately analyzed by MALDI-TOF (Fig. 2). The relative higher intensity of peaks with lower m/z in Fig. 2B compared to A is due to the interference of matrix background. Because the absolute peptide extract, including CYC, is lower in the tumor (Fig. 2B) compared to the normal (Fig. 2A), the relative intensity of the background seems higher. Peak 860.7 (Figs. 2A and B) is a good indicator of matrix noise, as it is not 18O labeled (Fig. 2B). Generated spectra were inspected and three predominant CYC peptide peaks were selected (see Section 2) (Fig. 2, Insets A1, A2, A3, B1, B2, and B3). The relative amount of peptides extracted from normal and adjacent tumor FFPE brainstem regions was estimated using the CYC peaks as internal control (see Section 2). To account for protein concentration differences, protein ratios generated by ZoomQuant were normalized to the internal control (horse CYC) following ZoomQuant analyses (see Section 2). Furthermore, spectra generated by MALDI-TOF were inspected to assess the efficiency of 18O labeling. Majority of inspected peaks (>95%) showed 4-Da shift due to incorporation of two 18O molecules as well as 2-Da shift due to the incorporation of a single 18O molecule. ZoomQuant software was used to account for single as well as double incorporation of 18O as described below.

3.2 Proteome profiling and identification of differentially expressed proteins in BSG FFPE tissues

Once efficient peptide extraction and isotopic labeling were confirmed, labeled and unlabeled sample pairs were mixed at 1:1 ratio and subjected to LC-MS/MS analysis as described in Section 2. To increase proteome coverage, we increased the peptide elution time from the most commonly used 90 to 240 min. This significantly increased the number of identified peptides when compared to samples that were processed for only 90 min (data not shown). We performed two sets of experiments, one, in which peptides extracted from normal sections were labeled with 16O and tumor with 18O and a reverse experiment, where peptides from tumour sections were labeled with 16O while peptides from normal section were labeled with 18O. Figure 3 shows a based peak chromatogram and a zoom scan of co-eluting pairs of labeled and unlabeled peptides with m/z of 795.8 and 793.8, respectively. These doubly charged ions show a 4-Da mass difference, indicating incorporation of two 18O atoms. A low intensity peak, however, was detected between the two isotopic clusters of the peptide pairs, corresponding to species with single 18O atom incorporation. We used ZoomQuant software, developed by Halligan and colleagues [25], to measure ratios between labeled and unlabeled peptides (Fig. 4). Briefly, after protein identification using SEQUEST search engine, both identified peptide list and the raw MS run were uploaded into the ZoomQuant software where zoom scans were extracted and ratios between labeled and unlabeled peptide pairs were measured (Fig. 4). Ratios were determined from the peak areas of isotope clusters of the labeled and unlabeled peptide pairs (Fig. 4) using the following equation L/U = [A1 + A2]/A0 where A0 represents the cluster area of the unlabeled peak, A1 the cluster area of the peak with a single 18O incorporation and A2 with a double 18O incorporation. Obtained ratios were then normalized to the ratios of internal standard (horse CYC) as described below.

Figure 3.

Figure 3

LC trace of a mixture of total extracted peptides from normal and control tissue. Shown is the LC trace of peptides eluted during a 4-h run. Peptides from tumor and normal areas were extracted, labeled by 18O and 16O, respectively, then mixed and analyzed by LC-MS/MS as described in Section 2. Inset shows a co-eluted labeled and unlabeled peptide pair generated from the same protein that is present equally in tumor and normal tissue.

Figure 4.

Figure 4

Schematic representation of 18O-labeled and unlabeled peaks by ZoomQuant software. Shown are 18O-labeled and unlabeled peaks for different peptides visualized by ZoomQuant software. A doubly charged peptide (GESGPSGPAGPTGARR) of collagen alpha 1(I) (Swiss-Prot ID P02452) (A) and a doubly charged peptide (LADVYQAELR) of glial fibrillary acidic protein (Swiss-Prot ID P14136) (C) were measured to be 1.8- and 2.8-fold up-regulated, respectively, in tumor versus normal. Reverse experiments, in which normal tissue was labeled with 18O, confirmed the observation (B and D). ZoomQuant software uses both single and double incorporation of 18O atoms in a given peptide to calculate the ratios between 16O- and 18O-labeled samples using the following equation: L/U = [A1 + A2]/A0 where A0 represents the cluster area of the unlabeled peak, A1 the cluster area of the peak with a single 18O incorporation and A2 with double 18O incorporation (D).

3.3 Analyses of identified proteins

The combined analyses of both forward and reverse 18O labeling identified and quantified 188 proteins. Initially, labeled and unlabeled peptide ratios (tumor to normal) levels were normalized to internal standard horse CYC and then relative expression levels of each protein was calculated (see Section 2). Our analyses showed that the expression levels of 134 proteins (71%) were not changed while 54 proteins (29%) showed up-regulation of ≥1.5-fold in tumor compared to the normal brain tissue and these are listed in Table 1. We then inspected this list for known markers of tumor formation, metastasis and cell invasion. We found that 15 (28%) of these 54 differentially expressed proteins have been previously reported as molecules with a potential role in tumor formation (Table 1A) and six of these proteins were indicative of angiogenesis or associated with various tumor types (Table 1B) while the remaining 33 could be potential novel components of pediatric BSG (Table 1C).

Table 1.

Potential pediatric brainstem glioma biomarkers

Protein gi SP_ID AverageFC Standarderror Peptidecount
A. Proteins previously associated with adult tissue/cell lines of supratentorial gliomas
Glial fibrillary acidic protein, astrocyte 121135 P14136 2.86 1.00 25
Vimentin 55977767 P08670 2.64 1.10 17
Heat-shock protein beta-1 19855073 P04792 2.56 0.01 2
Peroxiredoxin-6 1718024 P30041 2.45 1.10 17
Tubulin beta-2C chain 55977480 P68371 2.41 0.10 2
Neurofilament light polypeptide 62511894 P07196 2.13 0.45 6
Dihydropyrimidinase-related protein 2 (DRP-2) 3122051 Q16555 2.07 1.10 3
Superoxide dismutase 2 134665 P04179 1.97 0.75 2
Annexin A5 113960 P08758 1.90 0.10 6
Glucose-6-phosphate isomerase 17380385 P06744 1.86 0.20 2
Fructose-bisphosphate aldolase C 113613 P09972 1.82 0.56 5
Serum albumin precursor 113576 P02768 1.74 0.02 7
Alpha crystalline B chain 117385 P02511 1.68 0.28 2
L-lactate dehydrogenase B chain 126041 P07195 1.67 0.20 2
Pyruvate kinase isozymes M1/M2 20178296 P14618 1.64 0.45 6
B. Proteins previously associated with angiogenesis or various tumor types
Periostin precursor 93138709 Q15063 2.42 0.16 3
Hemoglobin subunit alpha 57013850 P69905 2.26 0.25 4
Hemoglobin subunit beta 56749856 P68871 1.98 0.95 2
Fibrinogen beta chain precursor 399492 P02675 1.96 0.43 1
Mimecan precursor 129078 P20774 1.71 0.13 2
Ferritin light chain 120523 P02792 1.66 0.50 2
C. Novel potential biomarkers associated with pediatric brainstem gliomas
α1-Collagen type VI 125987811 P12109 4.56 1.90 2
Transgelin 3123283 Q01995 3.42 1.40 4
Collagen alpha-2(VI) chain precursor 125987812 P12110 2.92 1.40 3
Ras-related protein Rap-1A precursor 51338596 P62836 2.80 1.40 1
ADAMTS-12 precursor 17366354 P58397 2.80 0.62 2
Spectrin beta chain, brain 1 116242799 Q01082 2.73 0.45 2
Keratin, type I cytoskeletal 10 547749 P13645 2.49 0.55 1
Myelin transcription factor 1-like protein 76789661 Q9UL68 2.46 1.20 2
Excitatory amino acid transporter 1 1169458 P43003 2.35 0.45 2
ATP synthase subunit beta, mitochondrial 114549 P06576 2.34 0.42 5
Creatine kinase B-type 125294 P12277 2.33 1.19 4
α3-Collagen type VI precursor 5921193 P12111 2.30 0.65 11
Heat shock protein HSP 90-alpha (HSP 86) 92090606 P07900 2.29 0.64 1
Collagen alpha-2(I) chain precursor 124056488 P08123 2.28 0.66 4
Spectrin alpha chain, brain 94730425 Q13813 2.23 0.30 3
Histone H2B type 1-L 7387743 Q99880 2.21 0.69 2
Tropomyosin alpha-4 chain 54039746 P67937 2.18 0.90 2
Myelin basic protein (MBP) 17378805 P02686 2.17 0.07 5
Protein S100-B 134138 P04271 2.16 0.45 2
Histone H2A type 2-A (H2A.2) 74757558 Q6FI13 2.07 0.81 5
Tubulin beta-4 chain 75075849 Q4R4X8 2.03 0.90 6
Tubulin alpha-3/alpha-7 chain 135418 P05214 1.97 0.15 2
Transgelin-2 586000 P37802 1.97 0.34 2
Alpha-actinin-1 46397817 P12814 1.91 0.38 2
Actin, alpha skeletal muscle 55976646 P68139 1.89 0.64 4
Collagen alpha-1(I) chain precursor 124056487 P02452 1.85 0.60 4
2,3″-cyclic-nucleotide 3″-phosphodiestera″ 1705945 P09543 1.70 0.01 2
Ferritin light chain 120523 P02792 1.66 0.51 2
Elongation factor 1-alpha 1 55584035 P68104 1.64 0.20 2
Tubulin alpha-6 chain 20455322 Q9BQE3 1.61 0.10 4
Plectin-1 (PLTN) 134044255 Q15149 1.61 0.01 2
Keratin, type II cytoskeletal 1 1346343 P04264 1.60 0.14 7
Glyceraldehyde-3-phosphate dehydrogenase 120649 P04406 1.60 0.31 4

4 Discussion

We present the first proteome profile of childhood BSG, thus providing new insight into the molecular pathogenesis of this disease. BSG differentially expressed proteins can be potential biomarkers and molecular signatures of disease progression for possible therapeutic intervention. Importantly, because the etiology of pediatric brainstem glioma has remained unknown, primarily due to the lack of viable tissue for experimental purposes, we describe a technique that may be utilized to study BSG from existing archived specimens, as well as more broadly for all other cancers. Given that FFPE tumor samples are routinely archived and accessible throughout the world, our goal was to develop methodologies to extract and quantify peptides from FFPE samples.

In this study, the largest amount of cells recovered from tumor, were neoplastic astrocytes. However, because of the heterogeneous nature of astrocytomas, other cell types that are an integral part of the tumor, such as the proliferating endothelial cells of vessels and to a much lesser extent, reactive stromal macroglial cells, normal glial cells and neurons, were necessarily present as well. It is important to note that we did not use LCM to recover the cells. Our aim was to recover the most representative area of tumor with the entire intrinsic cellular components in order to obtain the most robust protein signature specific to brainstem astrocytomas given that tumors are molecularly highly heterogeneous and thus random LCM may likely give an erroneous signature based on sampling alone. The same technique was used to recover the adjacent normal area of the brainstem not involved by tumor. The adjacent areas are composed of all the normal elements anatomically present in that specific area of the brain, including neurons, astrocytes, oligodendrocytes and macroglial cells.

Extracting intact proteins from FFPE samples involves removal of the paraffin wax and the reversal of formalin-induced cross-links. Wax removal is accomplished by a combination of heating, xylene and ethanol treatment. In 1991, Shi and colleagues [17] described a method for retrieval of antigens from FEPE tissues through microwave heating in the presence of metal solutions. The molecular mechanism of formalin fixation, wax removal and cross-link reversal of FFPE samples is well studied for antigen retrieval in immunohistochemical assays [16–20]. However, the majority of our BSG FFPE samples are between 10–25 years old. These tissue samples have been processed by different pathologists and possibly by different methods (e.g. fixation period from days to months). To address these issues, we combined and slightly modified existing protocols [16, 17, 20–23, 26, 27] and tested the feasibility of extracting proteolytically digested peptides for analysis and protein identification rather than extracting whole intact proteins. The concept of proteolytic digestion is that the enzyme (trypsin in this case) has access to cleavage sites on cross-linked proteins resulting in the easy release of peptides from the paraffin-embedded tissue. Here, we used a stepwise process in which de-waxed tissues were rehydrated in decreasing ethanol gradient followed by a prolonged tryptic digestion period (Fig. 1).

To ensure the validity of our findings we checked the yield of peptide extraction as well as the efficiency of 18O labeling at each step of our methodology. Since normal and tumor areas were not equal in size, protein measurement for each fraction was required. However, concentration measurements were hindered by the fact that we primarily dealt with peptides and not proteins. Although there are methods by which peptide concentrations can be measured (e.g. modified Bradford assays), in our experience, these methods either lack sensitivity or require a large amount of peptides. To address this issue, we spiked each sample with 2 µg of horse CYC (Swiss-Prot ID P00004), which served as internal standard. We chose horse CYC because it generates several unique tryptic peptides that are different in mass from tryptic peptides of human CYC. Since tryptic digestion of horse CYC was performed simultaneously with tryptic digestion of FFPE tissue, we were able to both test for the efficiency of 18O incorporation and measure the relative peptide concentration in each sample. Furthermore, to strengthen our approach, we performed reverse labeling experiments, in which tryptic peptides of normal brainstem sections were labeled with 18O, and compared to 16O-labeled peptides of tumor regions. Therefore, only proteins for which the expression levels were confirmed in both directions were retained.

Using this stringent methodology, we identified 54 proteins that were up-regulated in pediatric FFPE BSG when compared to the normal tissue. Our data agree with previous microarray gene expression profiling studies that have used either brain tumor cell lines or tissue from adult supratentorial gliomas [28–30]. We found that 15 out of the 54 differentially expressed proteins have been reported in these previous studies as potential tumor biomarkers. Among these, vimentin, peroxiredoxin-6, SOD1, annexin V, GFAP, and DRP-2 have been shown to be up-regulated in adult glioblastoma multiform (GBM) [28–31] whereas tubulin beta-2C chain, neurofilament light polypeptide, and pyruvate kinase are known to be up-regulated in meduloblastoma cell lines [29, 32]. We also detected a high expression of angiogenesis- related proteins in BSG tumor compared to the normal tissue sections and these included hemoglobin subunits, which are an indication of angiogenesis, as well as periostin and mimecan precursors. Overexpression of periostin is known to enhance invasiveness and angiogenesis in oral cancer cell lines [33]. The remaining 33 proteins that were found to be uniquely up-regulated (≥1.5-fold) in pediatric BSG might be considered as potential novel biomarkers of this disease. Indeed, our analysis of the 54 identified proteins using the ingenuity pathway analysis (IPA) software clustered 16 of these proteins in a single cancer-related pathway. IPA generates pathways based on curated publications taking into account known protein interactions. Perhaps the most intriguing BSG candidates in this pathway are α3-collagen type VI precursor (2.3-fold up-regulated), α1-collagen type VI (4.6-fold up-regulation), and a-spectrin (2.2-fold upregulated), which are excreted or membrane-associated proteins. Furthermore, these mentioned proteins are known to be involved in extracellular matrix remodeling and tumor progression [34, 35] (Fig. 5).

Figure 5.

Figure 5

Pathway generated by the Ingenuity software depicting interaction between proteins, whose expression was altered in BSG versus normal area. Shown is biological pathway generated by Ingenuity software depicting interaction of 16 proteins upregulated in BSG. Protein names are abbreviated and fold changes are shown where applied. Proteins highlighted in dark gray are up-regulated (shown as numerical values) in tumor when compared to the normal brainstem region (score of 45). Protein-protein interactions are shown as follows: indirect (Inline graphic), acts on (Inline graphic), or (Inline graphic) direct interactions.

Our observation that 61% of identified proteins did not overlap with previous studies of gliomas and other brain tumors, implies that pediatric BSG has a distinct pathology compared to gliomas in other locations of the brain. Furthermore, the strengths of our study are in the direct analysis of pediatric BSG tissue, rather than surrogate cell lines or adult glioma tissue, and in our comparison of BSG to the normal brainstem of the same patient, versus a different individual, thereby eliminating two main confounding variables. Further investigation using this methodology with larger sample sizes will be necessary to confirm that these proteins are novel tumor biomarkers and potential therapeutic targets of childhood BSG.

Acknowledgments

The authors would like to thank Dr. Eric P. Hoffman for his support and discussion and Dr. Kristy Brown for assistance with LC-MS/MS analysis via Genetic Medicine Proteomics Core facility. This work was supported by the Isabella Kerr Molina Foundation for Brainstem Glioma Research and Children’s National Medical Center (CNMC) Intellectual and Developmental Disabilities Research Center (IDDRC) grant (P30HD40677).

Abbreviations

BSG

brainstem glioma

CYC

cytochrome C

FFPE

formalin fixed paraffin embedded

Footnotes

The authors have declared no conflict of interest.

References