Lcd Proteome Analysis

Oral Oncology (2005) 41, 183–199

http://intl.elsevierhealth.com/journals/oron/

Proteome-wide analysis of head and neck squamous cell carcinomas using laser-capture microdissection and tandem mass spectrometry Haven Bakera, , Vyomesh Patelb, , Alfredo A. Molinolob, Edward J. Shillitoec, John F. Ensleyd, George H. Yooe, Abelardo Meneses-Garcı´af, Jeffrey N. Myersg, Adel K. El-Naggarh, J. Silvio Gutkindb,*, William S. Hancocka,* a

Chemistry and Chemical Biology Department, Barnett Institute, Northeastern University, 341 Mugar Building, 360 Huntington Avenue, Boston, MA 02115, USA b Oral and Pharyngeal Cancer Branch, National Institute of Craniofacial and Dental Research, National Institutes of Health, 30 Convent Drive, Building 30, Room 212, Bethesda, MD 20892-4330, USA c Department of Microbiology and Immunology, SUNY College of Medicine, Syracuse, NY 13210, USA d Department of Internal Medicine, Wayne State University/Karmanos Cancer Center, 4201 St Antoine, Detroit, MI 48201, USA e Department of Otolaryngology—Head and Neck Surgery, Wayne State University, 5E University Health Center, 4201 St Antoine, Detroit, MI 48201, USA f Department of Pathology, National Institute of Cancerology, Mexico. Av. San Fernando 22, Mexico, D.F. g Department of Head and Neck Surgery, 1515 Holcombe Blvd, U.T.M.D. Anderson Cancer Center, Houston, TX 77030, USA h Department of Pathology, 1515 Holcombe Blvd, U.T.M.D. Anderson Cancer Center, Houston, TX 77030, USA Received 2 June 2004; accepted 18 August 2004

KEYWORDS Oral cancer; Microdissection; Proteome; Biomarkers; Drug targets; Mass spectrometry

Summary Remarkable progress has been made to identify genes expressed in squamous cell carcinomas of the head and neck (HNSCC). However, limited information is available on their corresponding protein products, whose expression, post-translational modifications, and activity are ultimately responsible for the malignant behavior of this tumor type. We have combined laser-capture microdissection (LCM) with liquid chromatography–tandem mass spectrometry (LC–MS/MS) to identify proteins expressed in histologically normal squamous epithelium and matching

* Corresponding authors. Tel.: +1 301 496 6259; fax: +1 301 402 0823 (J.S. Gutkind); tel.: +1 617 373 4881; fax: +1 617 373 2855 (W.S. Hancock). E-mail addresses: [email protected] (J.S. Gutkind), [email protected] (W.S. Hancock).   These authors contributed equally to this work.

1368-8375/$ - see front matter Published by Elsevier Ltd. doi:10.1016/j.oraloncology.2004.08.009

184

H. Baker et al. SCC. The protein fraction from approximately 10,000–15,000 normal and tumor cells was solubilized, digested with trypsin, and the resulting peptides were analyzed by LC–MS/MS. Database searching of the resulting sequence information identified 30–55 proteins per sample. Keratins were the most abundant proteins in both normal and tumor tissues. Among the proteins differentially expressed, keratin 13 was much lower in tumors, whereas heat-shock (Hsp) family members were highly expressed in neoplastic cells. Wnt-6 and Wnt-14 were identified in both normal and tumor tissues, respectively, and placental growth factor (PIGF) was detected only in tumors. Immunohistochemical analysis of HNSCC tissues revealed lack of keratin 13 in tumor tissues, and strong staining in normal epithelia, and high expression of Hsp90 in tumors. Our study, by combining LCM and proteomic technologies, underscores the advantages of this approach to investigate complex changes at the protein level in HNSCC, thus complementing existing and emerging genomic technologies. These efforts may likely result in the identification of new biomarkers for HNSCC that can be used to diagnose disease, predict susceptibility, and monitor progression in individual patients. Published by Elsevier Ltd.

Introduction Annually, it is estimated that there are close to 500,000 cancer-related deaths in the United States alone, and of these approximately 13,000 are attributed to squamous cell carcinomas of the head and neck (HNSCC), making it the sixth most common cause of cancer deaths.1 Even though risk factors for HNSCC, such as the use of tobacco and alcohol, are well documented, a distinctive lack of suitable pre-malignant markers for early detection and risk assessment is clearly reflected by the fact that more than 50% of all HNSCC patients have advanced disease at the time of diagnosis.2–4 Indeed, the five year survival rate of HNSCC patients is in general poor, less than 50%, and the prognosis of the advanced HNSCC cases have not changed much over the past three decades.5 This limits the treatment options and renders management of HNSCC extremely challenging.6 Thus, the ability to identify and confidently predict malignant progression of HNSCC lesions will result in a reduction in mortality, by aiding in early diagnosis and treatment of this disease. Expression profiling using various microarray platforms, and large-scale cDNA sequencing projects, such as CGAP (cancer genome anatomy project), have led to a plethora of publicly available information on gene transcripts, which has proven to be fundamental for research efforts related to understanding both human biology and disease states.7–9 Despite this, there is only limited information available on the gene products, currently

estimated to be over 1 million proteins in a single cell, which play vital roles in most key cellular processes.10 Until recently, the analysis of a cell proteome using two-dimensional gels (IEF and SDS-PAGE) and mass spectrometry, was deemed technologically challenging.11 For instance, improved instrumental advances and the coupling of HPLC to electrospray mass spectrometry combined with the rapid growth in genomic databases amenable to searching with mass spectrometry data, now affords the opportunity to develop highthroughput proteomic approaches to identity minute amounts (typically femtomoles) of proteins present in complex samples.10,12,13 In that context, comparative analysis of the proteome in disease and normal cells, selectively procured by the use of laser-capture microdissection (LCM), is a critical step in the validation of the results because of the inherent clonal heterogeneity of most human cancers and the presence of host cells (fibroblast, endothelial and inflammatory cells).14 In this study, we have used LCM to isolate 10,000–15,000 normal and tumor epithelial cells from clinical samples of HNSCC, combined with mass spectrometry, to explore the feasibility of establishing a pattern of expressed cancer-related proteins for HNSCC. Our findings indicate that these approaches generate large proteomic datasets from minimal clinical samples that is likely to lead to the identification of novel HNSCC protein biomarkers. Indeed, some of the emerging protein information has already provided evidence of the expression of molecules that might be involved in tumor

Proteome-wide analysis of head and neck cancer progression, as well as clinically useful markers defining the margins of the neoplastic lesions.

Materials and methods

185 for 10 s and incubation at 65
C for approximately 3 h. The extraction solution was then recovered by centrifugation (14,000 rpm) for 2 min, pooled and stored at
80
C until ready to be processed for analysis.

Tissue samples

Preparation of protein extracts

Histologically squamous mucosa and tumor specimens from primary resected HNSCC from patients who provided written informed consent for planned studies and approved by Institutional Review Board, were immediately harvested by a head and neck pathologist, embedded in OCT (Tissue Tek compound, Sakura Finetechnical, CA) and stored at -70
C until use. Eight-micrometer cyrosections were cut on to standard RNAase free glass slides, which were stored at 70
C and used for downstream applications (see below) within a two-week period. Prior to use, all tissue sections were stained with hematoxylin and eosin, and confirmed by a pathologist (A.A.M) as being either dysplasia, malignant, invasive or normal epithelium (nonmalignant).

The protocol used to prepare protein extracts for mass spectrometry was adopted from the one used by Zhang et al.16 and the reagents used included sequence grade trypsin (Promega, Madison, WI), dithiothreitol (DDT), iodacetamide, ammonium bicarbonate (NH4HCO3), and formic acid (ICN Biomedical Inc., Aurora, OH). HPLC grade solvents (methanol, chloroform and acetonitrile) were purchased from Fisher Scientific (Hanover Park, IL). Briefly, the frozen protein extracts were thawed to room temperature, followed by denaturation (25 mM DDT) and alkylation (300 mM iodoacetamide). The salts and detergent were then removed from the samples by methanol/chloroform precipitation. After removal of the supernatant, the protein precipitates were rinsed once in methanol, air dried, and dissolved in 15 lL of 100 mM NH4HCO3 at pH 8.0, with agitation for up to 45 min. Dissolved proteins were digested with trypsin (1:20 w/w) at 26
C initially for 12 h, and then fresh trypsin was added at the same ratio and the digestion continued for an additional 12 h. The final volume of each sample was 17 lL.

Laser-capture microdissection (LCM) and protein extraction of HNSCC cells Frozen tissue section slides were stained prior to LCM, by briefly fixing in 70% ethanol (30 s), washed in purified dH2O, placed in MayerÕs hematoxylin for 30 s and subsequently rinsed in 70% ethanol, the slides were counterstained with eosin (10 s) and dehydrated with 95–100% ethanol and SAFECLEAR II (xylene substitute; Fisher Diagnostics, Middletown, VA, USA), and thoroughly air dried. For LCM, stained uncovered slides were viewed and after locating the cells of interest, a CapSure LCM Cap (Arcturus, Mountain View, CA, USA) was placed over the target area and pulsed with laser to adhere cells to the cap, which after sufficient capture (5000 cells) was transferred to a 0.5 mL sterile microfuge tube for immediate processing. For each sample, approximately 10,000– 15,000 cells were captured on multiple caps, which were subsequently processed for protein extraction. The procedure of protein extraction from CapSure caps was essentially as described.15 Briefly, 30–50 lL PicoPure (Arcturus) protein extraction buffer was placed in each tube and with the enclosed caps, the tubes were inverted ensuring the buffer was adequately dispersed to cover the surface of the caps, followed by vortexing

Liquid chromatography (LC)–tandem mass spectrometry (MS/MS) analysis of samples LC–MS/MS was performed on a ThermoFinnigan Deca-XP coupled to a Surveyor LC system (ThermoFinnigan, San Jose, CA). Fifteen microliters of the analyte solution was injected on a Thermo-Hypersil 100 · 0.18 mm C18 reversed phase microcolumn. A 210 minute 0.1% formic acid, ACN/0.1% formic acid, H2O gradient was followed by analysis on the orthogonal microspray ion trap mass spectrometer. The flow rate was 150 lL a minute, and after splitting this was maintained at 1.5 lL a minute. The dynamic exclusion mode on the mass spectrometers was used. To obtain peptide fragmentation spectra, each MS spectra was followed by 3 MS/MS scans and for analysis, precursor ion +/
2 Da was excluded for 1.5 min if it was analyzed twice in the previous 30 s, and normalized collision energies were set at 35%.

186

Database search and protein identification Data analysis and protein identification of the emerging spectra was performed with the TurboSEQUEST search engine using default parameters and the SWISS-PROT human protein database, and the bioinformatics of the resulting information was as previously described.17 Briefly, the algorithm compares certain parameters for instance, similarities between theroretical peptides derived from the database and those from experimental MS/MS scans, and subsequently assigns a unified ranking score (combination of delta CN and X corr without a bias against small peptides), which requires a minimum of 2400 to discriminate between background noise and true identification. The search parameters included modification of cysteine by carbamidomethylation. The assignments of lowlevel peptide sequences were then manually confirmed by comparing the acquired MS/MS spectra to the theoretical fragmentation patterns. The presence of multiple peptides of differing mass and sequence of the same protein also indicated favorably to the identity of the protein.

Immunohistochemistry Archival HNSCC tissue sections were deparaffinized in Safeclear II (Fisher Scientific, USA) and hydrated through graded alcohols, distilled water and PBS 1x. Cytokeratin 13 was retrieved by incubating with 0.25 mg/mL trypsin (Invitrogen Corporation, Carlsbad, CA, USA) in PBS, for 30 min at 37
C. Endogenous peroxidase activity was quenched by incubation in 3% hydrogen peroxide in 96% alcohol. Each incubation step was performed and followed by three sequential washes in PBS for 5 min each. Slides were incubated in blocking solution (5% horse serum) for 30 min at room temperature and reacted with the indicated primary antibodies (Cyotkeratin 13, Novacastra Laboratories Ltd, Newcastle upon Tyne, UK;Hsp90, Stressgen Biotechnologies, Victoria, Canada) diluted 1:100 or 1:40, respectively, in blocking solution at 4
C, overnight. Sections were then washed with PBS, and incubated with biotinylated secondary antibody (Vector Laboratories, Burlingame, CA, USA) for 30 min, washed again and reacted with the ABC complex, prepared according to the manufacturer’s instructions (Vector Stain Elite, ABC kit, Vector Laboratories) for 30 min at room temperature. The peroxidase was visualized with 3,3-diaminobenzidine (Sigma FASTDAB tablets with metal enhancer, Sigma Chemical, St. Louis, MO, USA) as chromogen substrate and closely monitored the

H. Baker et al. staining under microscope. The reaction was stopped with distilled water. Cytokeratin 13, were counterstained with hematoxylin, washed in tap water, dehydrated and mounted with glass cover slips. Hsp90 slides were not counterstained, as a faint positive nuclear reaction was evident at developing time that may have been masked by hematoxylin.

Results HNSCC patient sets For the proteomic analysis of HNSCC, we initially selected five cases of matched histologically normal squamous epithelium and carcinoma from each patient. As a criterion for selection, we elected to focus on SCC from the tongue, the most frequent anatomical location of the primary HNSCC lesions.18 As indicated in Table 1, these lesions were diagnosed as poorly (Case 1), moderate to well-moderate (Case 2–4) or well-differentiated (Case 5) tongue carcinomas. Normal mucosa was defined tissue with squamous epithelium histologically lacking hyperplastic or dysplastic features by light microscopic examination. Additionally, the age of the HNSCC patients who underwent resection ranged from 48 to 74 years and all were of male gender, which correlated well with the patient population at the highest risk of developing this disease.5

Laser-capture microdissection (LCM) We have previously applied LCM to successfully isolate pure populations of normal and tumor cells from HNSCC tissues for the subsequent gene expression analysis.19 However, a similar application to interface LCM with the analysis of the proteome of specific cell populations could be limited primarily due to sample availability and the absence of in vitro amplification steps for protein identification. Indeed, to succeed, this approach is likely to depend on the use of highly sensitive protein detection systems, together with the accuracy of LCM for the isolation of specific cell types, and the quality and preservation of the tissue samples. To this end, LCM generally ensures approximately 95% purity of the correct cell population from tissues with optimum tissue integrity intact.19 Thus, before proceeding with the analysis, we first confirmed the quality and the histopathology of all the samples by microscopic visualization

Proteome-wide analysis of head and neck cancer Table 1

187

Clinical and histological information on HNSCC matched patient sets

Case

Tissue

Age

Sex

Site

Grade

1

N T N T N T N T N T

58

M

Tongue

74

M

Tongue

69

M

Tongue

68

M

Tongue

48

M

Tongue

Squamous epithelium Poorly differentiated Squamous epithelium Moderately differentiated Squamous epithelium Moderately differentiated Squamous epithelium Well-moderately differentiated Squamous epithelium Well differentiated

2 3 4 5

Each case consisted of non-malignant (N) squamous epithelium and tumor counterpart (T). Age and the sex of the HNSCC patients, tumor grade, and the site within the oral cavity that the tissues originated from, is indicated.

of frozen tissue sections (8 lm thickness), stained with hematoxylin-eosin. Oral squamous epithelium lacking hyperplastic and/or dysplastic cellular features were localized carefully for microdissection. Similarly, all tumor biopsies were assessed to be squamous cell carcinomas with variable histological grades, which ranged from poorly to well-dif-

ferentiated lesions. After this initial assessment, all the samples were considered suitable for LCM and proteome analysis. Successful microdissection of a representative sample (tumor) is depicted in Figure 1, whereby tumor cells once identified (a) are captured by laser onto caps (b) and after confirming the cell purity (c–d), processed for protein

Figure 1 Laser-capture microdissection of HNSCC: Tumor cryosections were fixed, stained with hematoxylin and eosin (H&E), dehydrated and analyzed under the microscope. Once tumor cells of interest were identified (a), microdissection was performed (b), and the cell purity confirmed by inspecting the area of tissue that underwent procurement (c) and the captured cells (d). Bar represents 100 lm.

188

H. Baker et al. more traditional techniques like 2-D gels.20 LC– MS/MS is now sensitive enough to analyze only a few cells, making it compatible with LCM and thus suitable for proteome-wide profiling of cancer tissues. To this end, protein extracts of cells microdissected from HNSCC patient sets underwent global proteolysis with trypsin, and the resulting complex peptide mixture were subjected to liquid chromatography, and the separated molecules were fed directly into ESI (electrospray ionization) and tandem mass spectrometer (MS/MS). Briefly, ESI ionizes peptides by passing the solution through a high-voltage nozzle. Analyzers arranged in tandem use radiofrequency and direct-current voltages to analyze ions based on their mass and charge. The resulting full scan mass spectra was collected in real time in a fully automated fashion

extraction. Using this procedure, cells from each tissue sample were rapidly captured on multiple caps.19

LC–MS/MS analysis of microdissected HNSCC cells A major challenge for proteomics research today, is the identification of all the proteins in a given biological system, as they exist in vivo. This is due primarily to the vast complexity and the dynamic range of proteins, and concurrent with techniques that lack the desired sensitivity of detection. Nonetheless, recent studies have indicated that the LC–MS/MS approach may significantly improve the dynamic range of protein detection and identification in comparison with

C:\Haven\oral_cancer_101003\run_8228N

10/11/200308:41:56PM

RT : 0.34-119.99 77.67

100 90 34.10

Relative Abundance

80

NL: 1.29E7 BasePeak MS run_8228N

54.77

44.41 47.00

70

25.89 26.12

60

56.39

36.98

58.37 61.45

50

78.24 30.11

68.14

40

20 0.66 1.94

10

19.07

90.24 94.01 97.17

117.92

82.95

66.64

24.30 22.55

30

76.30

105.79 108.37 111.08

21.52

13.45 15.12

0 10

20

30

40

50

60

70

80

90

100

110

Time(min) R.DAEEWFHAK.S

run_8228N #1288 RT: 35.25 AV: 1 NL: 1.39E5 T: + c ESI d Full ms2 [email protected] [ 145.00-1150.00] 817.2

100

Relative Abundance

688.2 465.0

80

502.2

60

818.2

915.2

187.0

20

689.2

509.5

40

218.1

631.0 316.0

445.0

799.5

558.4

736.5

887.2 1004.2

0 200

300

400

500

600

700

946.2

800

900

1000

1100

m/z

Figure 2 MS analysis of HNSCC: Whole cell protein extracts of LCM procured cells underwent global proteolysis with trypsin and the resulting peptide fragments were first separated by liquid chromatography, followed by electrospray ionization and analysis in a single MS mode, giving a mass spectrum of the peptides eluted with time (upper panel). Next, fragments underwent tandem mass spectrometric sequencing and identification of selected peptides by matching against protein sequence database (lower panel).

Proteome-wide analysis of head and neck cancer Table 2

189

Proteins expressed in LCM-procured normal oral epithelium

Proteins identified Keratin, type I cytoskeletal 13 Keratin, type I cytoskeletal 17 Keratin, type II cytoskeletal 6F Keratin, type II cytoskeletal 5 Keratin, type I cytoskeletal 14 Keratin, type II cytoskeletal 1 Keratin, type II cytoskeletal 6E Keratin, type I cytoskeletal 10 Keratin, type II cytoskeletal 4 Annexin I (lipocortin I) (calpactin II) (chromobindin 9) Actin, cytoplasmic 2 (gamma-actin) Keratin, type I cytoskeletal 19 Keratin, type II cytoskeletal 2 epidermal Serum albumin precursor Hemoglobin alpha chain Fibrinogen gamma chain precursor Keratin, type II cytoskeletal 6A Junction plakoglobin (desmoplakin III) IGS Annexin II (lipocortin II) (calpactin I heavy chain) Elongation factor 1-alpha 1 (EF-1-alpha-1) Heat shock 27 kDa protein (HSP 27) (stress-responsive protein 27) Spectrin beta chain, brain 1 (spectrin, non-erythroid beta chain 1) WNT-6 protein precursor Rod CGMP-specific 30 ,50 -cyclic phosphodiesterase alpha-subunit Hemoglobin beta chain Histone H2A.L (H2A/L) Collagen alpha 2(I) chain precursor Keratin, type II cytoskeletal 6C Histone H4 Histone H3.3 (H3.A) (H3.B) (H3.3Q) Keratin, type I cytoskeletal 9 Thioredoxin (ATL-derived factor) (ADF) (surface associated sulphydry) Keratin, type I cytoskeletal 15 Elongation factor 2 (EF-2) Calnexin precursor (major histocompatibility complex class I antigen) Hypothetical zinc finger protein KIAA0296 T-complex protein 1, delta subunit (TCP-1-delta) Splicing factor U2AF 65 kDa subunit Acylamino-acid-releasing enzyme (acyl-peptide hydrolase) Keratin, type I cytoskeletal 12 (cytokeratin 12) Natural resistance-associated macrophage protein 1 (NRAMP 1) Heat shock 70 kDa protein 1 (HSP70.1) (HSP70-1/HSP70-2) EZRIN (p81) (cytovillin) (villin 2) Serine/threonine protein phosphatase 2B catalytic subunit, alpha is Nonsyndromic hearing impairment protein 5 Triple functional domain protein (PTPRF interacting protein) C-terminal binding protein 1 (CTBP1) Desmoplakin (DP) (250/210 kDa) Lim domain kinase 2 (LIMK-2) 80 kDa MCM3-associated protein (ganp protein) Nucleolar phosphoprotein P130 Transaldolase Hypothetical protein KIAA0064 (HA1355) Propionyl-COA carboxylase alpha chain, mitochondrial precursor CTD-binding SR-like protein RA4

Samples 5 5 5 5 5 5 5 5 4 4 4 4 4 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 (continued on next page)

190

H. Baker et al.

Table 2 (continued) Proteins identified

Samples

Kinesin-like protein KIF3C Keratin, type II cytoskeletal 8 Histone deacetylase 4 (HD4) (HA6116) Interleukin-14 precursor (IL-14) Opioid binding protein/cell adhesion molec Transcription initiation factor TFIID 135 Jumonji protein NG,NG-dimethylarginine dimethylaminohydrol Eukaryotic translation initiation factor 3 Peripheral plasma membrane protein cask Insulin-like growth factor I receptor precu Myosin heavy chain, skeletal muscle, adult 2 Pregnancy zone protein precursor Alpha-1-antitrypsin precursor (alpha-1 protease inhibitor) Fibrinogen alpha/alpha-E chain precursor 14-3-3 protein sigma (stratifin) (epithelial cell marker protein 1) Cadherin-19 precursor Neuropeptide Y receptor type 4 (NPY4-R) (pancreatic polypeptide rec) Alpha-2-macroglobulin precursor Eosinophil peroxidase precursor (EPO) Ornithine decarboxylase antizyme 2 (ODC-AZ 2) (AZ2) Nitric-oxide synthase, brain (NOS, type I) Glutathione synthetase (glutathione synthase) Vitronectin precursor (serum spreading factor) (S-protein) Tenascin precursor (TN) (hexabrachion) Caspase-9 precursor (CASP-9) (ice-like apoptotic protease 6) Keratin, type II cytoskeletal 2 oral DNA topoisomerase I Keratin, type I cuticular HA3-II DNA-repair protein XRCC1 Neurofilament triplet H protein Surfeit locus protein 6 Desmocollin 3A/3B precursor Alpha-actinin 1 Peripheral plasma membrane protein cask Zinc finger protein 222 Cholinephosphate cytidylyltransferase A Versican core protein precursor (large fibroblast proteoglycan) Fukutin precursor (fukuyama-type congenital muscular dystrophy pro) Isoleucyl-TRNA synthetase, cytoplasmic (isoleucine–TRNA ligase) Trichohyalin Period circadian protein 1 (circadian pacemaker protein RIGUI) Developmentally regulated GTP-binding protein 1 (DRG 1) Protein tyrosine phosphatase, non-receptor type 14 Myosin VB (myosin 5B) Tubulin beta-4 chain (tubulin beta-III) RAS-related protein RAB-35 (RAB-1c) (GTP-binding protein ray) DNA ligase III (polydeoxyribonucleotide synthase [ATP])

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Proteins detected in combined Normal samples. Number of samples in which the protein was detected is indicated.

to identify the constituent peptides, followed by tandem mass spectrometry and sequencing of selected peptides in each time interval. This MS to MS–MS cycle spans approximately 0.2 s, enabling

a steady stream of peptide and fragment masses to be fed to a computer for off-line database search and protein identification. A representative ion chromatogram mass spectrum for one of the

Proteome-wide analysis of head and neck cancer Table 3

191

Proteins expressed in LCM-procured tumor squamous oral epithelium

Proteins identified in tumor Keratin, type II cytoskeletal 1 Keratin, type II cytoskeletal 6F Keratin, type I cytoskeletal 14 Keratin, type I cytoskeletal 17 Keratin, type II cytoskeletal 5 Keratin, type II cytoskeletal 6E Keratin, type I cytoskeletal 10 Serum albumin precursor Keratin, type II cytoskeletal 2 epidermal Actin, alpha skeletal muscle Glyceraldehyde 3-phosphate dehydrogenase, liver Hemoglobin alpha chain Actin, cytoplasmic 2 Fibrinogen gamma chain precursor Keratin, type II cytoskeletal 6A Keratin, type I cytoskeletal 9 Elongation factor 1-alpha 2 (EF-1-alpha-2) Desmoplakin (DP) (250/210 kDa) Tubulin alpha-4 chain Keratin, type I cytoskeletal 16 Keratin, type II cytoskeletal 3 Keratin, type II cytoskeletal 8 Heat shock protein HSP 90-beta (HSP 84) (HSP 90) Elongation factor 1-alpha 1 (EF-1-ALPHA-1) Heat shock 27 kDa protein (HSP 27) Histone H4 Heat shock 70 kDa protein 1 (HSP70.1) Pregnancy zone protein precursor Histone H2A.L (H2A/L) Annexin I (lipocortin I) (calpactin II) Hemoglobin beta chain Spectrin beta chain, brain 1 (spectrin, non-erythroid beta chain 1) Protein-tyrosine phosphatase-like N precursor Insulin receptor precursor (IR) Fatty acid-binding protein, epidermal (E-FABP) (psoriasis-associated) WNT-14 protein Zinc finger protein 43 (zinc protein HTF6) DNA topoisomerase I Cholinephosphate cytidylyltransferase A (phosphorylcholine transfer) Vacuolar ATP synthase subunit C (V-ATPase C subunit) T-cell differentiation antigen CD6 precursor (T12) (TP120) Vimentin Histone H3.3 (H3.A) (H3.B) (H3.3Q) Glutamate [NMDA] receptor subunit epsilon Collagen alpha 2(VI) chain precursor Camp and camp-inhibited CGMP 30 ,50 -cyclic Signal recognition particle 68 kDa protei Inositol polyphosphate 5-phosphatase OCRL Complement decay-accelerating factor precur Solute carrier family 2, facilitated glucos ACTG_human actin, cytoplasmic 2 Calgranulin A (migration inhibitory factor-related protein 8) Galectin-7 (HKL-14) Annexin II (lipocortin II) (calpactin I heavy chain) 14-3-3 Protein gamma (protein kinase C inhibitor protein-1) Ephrin-B2 precursor (EPH-related receptor tyrosine kinase ligand 5)

Samples 5 5 5 5 5 5 5 5 5 4 3 4 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 (continued on next page)

192

H. Baker et al.

Table 3 (continued) Proteins identified in tumor

Samples

Fructose-bisphosphate aldolase A (muscle-type aldolase) (lung cancer) Cortistatin precursor [contains: cortistatin-29;cortistatin-17] Calnexin precursor (major histocompatibility complex class I antigen) Peroxiredoxin 1 (thioredoxin peroxidase 2) Gamma-interferon inducible lysosomal thiol reductase precursor CENP-F kinetochore protein (centromere protein F) Cell division cycle 2-related protein kinase 7 RAS GTPASE-activating-like protein IQGAP1 (P195) Placenta growth factor precursor (PlGF) 40S ribosomal protein S26 Origin recognition complex subunit 2 Cleavage stimulation factor, 64 kDa subunit GMP synthase [glutamine-hydrolyzing] Fibrinogen gamma chain precursor Trans-1,2-dihydrobenzene-1,2-diol dehydrogenase Keratin, type I cytoskeletal 15 Lysosome-associated membrane glycoprotein 1 precursor (LAMP-1) Endothelial actin-binding protein (ABP-280) Alpha-2-macroglobulin precursor (ALPHA-2-M) 78 kDa glucose-regulated protein precursor (GRP 78) Myosin heavy chain, skeletal muscle, adult 2 Cyclic-AMP-dependent transcription factor ATF-6 Keratin, type I cytoskeletal 13 (cytokeratin 13) (K13) (CK 13) [MAS] Keratin, type II cytoskeletal 4 Keratin, type I cytoskeletal 19 (cytokeratin 19) Keratin, type I cuticular HA6 (hair keratin, type I HA6) Keratin, type I cuticular HA3-II (hair keratin, type I HA3-II) D-site-binding protein (albumin D box-binding protein) Down syndrome critical region protein 5 (Down syndrome critical reg) Alpha-1 catenin (cadherin-associated protein) (alpha E-catenin) P55-c-FOS proto-oncogene protein (cellular oncogene c-FOS) Interstitial collagenase precursor (matrix metalloproteinase-1) Serine/threonine protein phosphatase 2A, 72/130 kDa regulatory CENP-F kinetochore protein (centromere protein F) (mitosin) Inward rectifier potassium channel 2 Trifunctional enzyme beta subunit, mitochondrial precursor (TP-BETA)

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Proteins detected in combined tumor samples. Number of samples in which the protein was detected is indicated.

HNSCC cases is shown in Figure 2, whereby a complex mixture of tryptic digested peptides, were separated and a peptide for cytokeratin 13 was sequenced from its MS/MS fragmentation pattern. The data indicate that protein identification can be achieved from small sample material.

Identification of proteins expressed in HNSCC Although a list of candidate proteins expressed in both normal oral epithelium and tumor specimens can be obtained using 2-D gels, the enhanced sensitivity needed for analyzing LCM samples requires LC–MS/MS and the tryptic digestion of the

samples. This generates several dozen peptides per protein and leads to complex proteome profiling due to the large number of redundant peptides. Thus, data confidence to identify true proteins from background noise is primarily facilitated by the use of computational algorithms. For each sample analyzed in this study, a list containing the identity of numerous detected molecules was generated, but only those demonstrating high confidence (unified score >2400 and multiple peptide coverage) were selected.15,17 In this regard, between 21–53 distinct proteins were readily identified in each sample (data not shown). Although many proteins were noted to be common to most or all normal or tumor tissues, a significant number

Proteome-wide analysis of head and neck cancer Table 4

193

Protein frequency in normal (panel A) and tumor (panel B) oral epithelium

Panel A: Abundant proteins in normal Keratin, type I cytoskeletal 13 Keratin, type I cytoskeletal 17 Keratin, type II cytoskeletal 6F Keratin, type II cytoskeletal 5 Keratin, type I cytoskeletal 14 Keratin, type II cytoskeletal 1 Keratin, type II cytoskeletal 6E Keratin, type I cytoskeletal 10 Keratin, type II cytoskeletal 4 Annexin I (lipocortin I) (calpactin II) (chromobindin 9) Actin, cytoplasmic 2 (GAMMA-ACTIN) Keratin, type I cytoskeletal 19 Keratin, type II cytoskeletal 2 epidermal Serum albumin precursor Hemoglobin alpha chain Fibrinogen gamma chain precursor Keratin, type II cytoskeletal 6A Junction plakoglobin (desmoplakin III) IGS Annexin II (lipocortin II) (calpactin I heavy chain) Elongation factor 1-alpha 1 (EF-1-ALPHA-1) Heat shock 27 kDa protein (HSP 27) (stress-responsive protein 27) Spectrin beta chain, brain 1 (Spectrin, non-erythroid beta chain 1) WNT-6 protein precursor ROD cGMP-specific 30 ,50 -cyclic phosphodiesterase alpha-subunit Hemoglobin beta chain Histone H2A.L (H2A/L) Collagen alpha 2(I) chain precursor Keratin, type II cytoskeletal 6C Histone H4 Histone H3.3 (H3.A) (H3.B) (H3.3Q) Keratin, type I cytoskeletal 9 Heat shock 70 kDa protein 1 (HSP70.1) (HSP70-1/HSP70-2) Desmoplakin (DP) (250/210 kDa) Glyceraldehyde 3-phosphate dehydrogenase, liver Keratin, type II cytoskeletal 8 Panel B: Abundant proteins in tumor Keratin, type II cytoskeletal 1 Keratin, type II cytoskeletal 6F Keratin, type I cytoskeletal 14 Keratin, type I cytoskeletal 17 Keratin, type II cytoskeletal 5 Keratin, type II cytoskeletal 6E Keratin, type I cytoskeletal 10 Serum albumin precursor Keratin, type II cytoskeletal 2 epidermal Actin, alpha skeletal muscle Hemoglobin alpha chain Glyceraldehyde 3-phosphate dehydrogenase, liver Actin, cytoplasmic 2 Fibrinogen gamma chain precursor Keratin, type II cytoskeletal 6A Keratin, type I cytoskeletal 9 Elongation factor 1-alpha 2 (EF-1-alpha-2)

Samples 5 5 5 5 5 5 5 5 4 4 4 4 4 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 1 Samples 5 5 5 5 5 5 5 5 5 4 4 3 3 3 3 3 3 (continued on next page)

194

H. Baker et al.

Table 4 (continued) Desmoplakin (DP) (250/210 kDa) Tubulin alpha-4 chain Keratin, type I cytoskeletal 16 Keratin, type II cytoskeletal 3 Keratin, type II cytoskeletal 8 Heat shock protein HSP 90-beta (HSP 84) (HSP 90) Elongation factor 1-alpha 1 (EF-1-ALPHA-1) Heat shock 27 kDa protein (HSP 27) Histone H4 Heat shock 70 kDa protein 1 (HSP70.1) Pregnancy zone protein precursor Histone H2A.L (H2A/L) Annexin I (lipocortin I) (calpactin II) Hemoglobin beta chain Spectrin beta chain, brain 1 (spectrin, non-erythroid beta chain 1) WNT-14 protein Histone H3.3 (H3.A) (H3.B) (H3.3Q) Annexin II (lipocortin II) (calpactin I heavy chain) Keratin, type I cytoskeletal 13 (cytokeratin 13) Keratin, type I cytoskeletal 19 (cytokeratin 19) Collagen alpha 2(VI) chain precursor Histone H3.3 (H3.A) (H3.B) (H3.3Q)

3 3 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1

Number of samples in which the protein was detected is indicated. Proteins highlighted in italic font are more frequent in normal oral epithelium, whereas those highlighted in bold font are more frequent in tumor epithelial cells.

of proteins identified were unique to individual samples. In this manner, approximately 90–105 unique proteins were identified in all normal and tumor samples. Table 2 shows a partial list of the proteins identified based on a unified score >2400, and the frequency of detection in normal samples (n = 5). In this group, the keratins were readily detected in all five normal samples. These include keratin 1, 2, 4, 5, 6, 10, 13, 17 and 19. Some structural (actin, junction plakoglobin, collagen), serum (serum albumin, hemoglobin), histone (H2A/L, H3, H4), were also identified. Proteins of interest detected in 2/5 samples, include Wnt-6, Annexin I and II, Hsp27 and elongation factor 1 alpha 1. Hypothetical Zinc finger protein (KIAA0296), Jumonji, Pregnancy Zone protein precursor, 14-3-3 sigma, caspase 9 precursor, DNA Topoisomerase 1, DNA repair protein XRCC1 and Rab 35 were present in at least 1/5 normal samples. In the tumor samples (Table 3), the keratins were again well represented, with the exception of keratin 13 and 19 that were much less expressed than in normal squamous tissue. There was some indication that the keratin 6 isoforms (6C) may be differentially expressed. Furthermore, Hsps were highly detected in the tumor cells (Hsp27, 70, 90). Elongation factor 1 alpha 2 and desmoplakin were present in multiple tumor samples (3/5). Additional proteins of

interest include Wnt-14, psoriasis-associated fatty acid binding protein, 14-3-3 gamma, cdc2 related protein kinase 7, Ras GTPase activating like protein, placenta growth factor precursor (PIG), alpha 1 catenin, c-Fos, MMP-1 and mitosin, which were detected in one of five tumor samples. Frequency of selected proteins detected in all normal and tumor samples are summarized in Table 4 (panels A and B), which also indicates the differential expression of keratin (6C, 13, 19) and Hsp (70, 90) family members.

Detection of cytokeratin 13 and Hsp90 by immunohistochemistry in archival HNSCC tissues As our previous data indicated that a sub-group of proteins might be differentially expressed in HNSCC tissue samples, we opted to confirm the presence or absence of cytokeratin 13 and Hsp90 in a representative set of HNSCC tissues. Because frozen sections offer poor morphology for immunohistochemistry, sections from paraffin-embedded archival HNSCC tissues were used. As seen in Figure 3a, cytokeratin 13 was expressed in normal epithelia, and the staining was preferentially located within the upper layers of the squamous epithelium

Proteome-wide analysis of head and neck cancer

195

Figure 3 Detection of keratin 13 and Hsp90 by immunocytochemistry in archival HNSCC. Archival oral squamous cell carcinoma paraffin sections were assessed for cytokeratin 13 (a) and Hsp90 (b), as described in materials and methods section . Expression of cytokeratin 13 is shown to be present in normal epithelia (left panel), while tumor cells show reduced levels (middle panel) and H&E staining indicates the boundary between the tumor and stoma (fl), and distal to non-malignant normal epithelia (right panel). Expression of Hsp90 protein was undetectable in normal tissues (left panel), while elevated levels were readily observed in cells (right panel). Bar represents 100 lm.

(left panel). In malignant tumors the expression of cytokeratin 13 is lost; adjacent non-malignant epithelial cells were positively stained (middle panel). The highlighted area (arrows) distinguishes the margin of the tumor from the adjacent stroma,

and the distal non-malignant epithelium (middle and right panels). For Hsp90 (Fig. 3b), non-neoplastic squamous epithelium (left panel) gave a uniformly mild cytoplasmic staining with scattered nuclear reaction; in malignant tumors (right panel)

196 the immunoreactivity was exclusively cytoplasmic and moderate to strong in intensity.

Discussion Direct and rapid analysis of low abundance proteins in complex mixtures by mass spectrometry is a compelling approach to comprehensively identifying protein components. It provides a list of actual proteins present in a purified complex instead of a descriptive visualization of the components that must be individually identified.21 Typically, 1D or 2D gel electrophoresis, a time- and laborintensive process with limited molecular mass or pI ranges, is used to resolve complicated protein mixtures into individual bands or spots.11 Peptides from digested proteins must be recovered from the stained gel or an electroblot of the gel. Automation of this process requires expensive robotics to isolate and process the spots. Directly identifying proteins from the digest of a complex sample bypasses the potential limitations of gel electrophoresis, including protein insolubility, limited fractionation ranges, and limited recoveries of material, and provides the ability to perform proteomic analyses on samples containing a few thousand cells that could open the possibility of performing individual patient studies.22,23 However, the identification of proteins in a few cells has been often associated with technical challenges such as the solubilization, extraction and separation of whole cellular proteins. A recent improvement in buffers and separation techniques, such as liquid chromatography coupled directly to MS/MS, makes it possible to generate high-quality data from a very little starting material. In this regard, a recent study using LCM to collect a small number of homogeneous cells for proteomic analysis by the LC–MS/MS was evaluated, which helped establish extraction procedures enabling the subsequent solubilization of the majority of the protein complexes in a single step.15 Although the use of cultured cells can serve as a model system for clinical studies, the biggest challenge in cancer proteomics is to unravel the molecular complexity of the tissue microenvironment, and thus the proteomic analysis of cells cultured in vitro may not truly reflect the expression pattern of cells in vivo.24 To this end, we opted to analyze a representative set of HNSCC patient samples, in order to identify proteins that may provide insights to the pathogenesis of the disease or potential bio-

H. Baker et al. markers. It is noteworthy that two-dimensional chromatography is typically used for separation, but for this study, all the tissue samples underwent a single dimension RPLC, primarily to maximize the recovery of small samples of complex protein mixtures, 1–5 lg of total protein, and to minimize contaminants from sample handling. Altogether, this novel method enabled the identification of approximately 91–103 proteins expressed in either normal or tumor tissues. From the proteins detected, an assessment was made to determine whether any of these molecules were differentially detected. Down-regulated proteins in tumors included several cellular structural molecules such as cytoskeletal keratins 13, 19, and 6c. These were all found in the normal samples but at a much lower levels and frequency in the tumor samples. Studies have shown that down-regulation of specific cytokeratins in squamous cell carcinomas correlates in part with the general loss of differentiation.25 Specifically, cytoskeletal keratins 13 and 19 are often less expressed in tumor than normal cells.26,27 Our findings of both keratins in the normal and not in corresponding malignant tissues support these studies. We further confirmed the expression of keratin 13 in archival tissues, and noted that protein levels were primarily confined to the upper layer of the normal squamous epithelia, and not to the tumor counterpart. These observations were further corroborated during gene array analysis of HNSCC samples from an independent study, in which we readily observed the presence of keratin 13 mRNA in normal samples, but its absence in tumors (data not shown). Although the mechanism leading to the loss in expression of keratin 13 in HNSCC is unclear, this protein might play a role in the HNSCC carcinogenesis, possibly by acting as a tumor suppressor gene product, and thus its level of expression may be relevant for the diagnosis and prognosis of this cancer type. Additionally, our findings suggest that keratin 13 (and/or19) might be used as a clinical marker, whereby the presence or absence of this protein could aid to define the surgical margin.28 Of the up-regulated proteins, the Hsps were the most notable in the tumor samples. Hsp70 and Hsp90 kDa both appeared in two tumor samples, but not in the controls. While the presence of Hsp may indicate a general response to biological stress experienced by the tumor cells, the precise role of these proteins in HNSCC is not clear.29,30 Interestingly, there are reports indicating that Hsp70 may promote proliferation and survival of oral tumor cells, possibly by interacting and inhibiting apoptosis mediators, such as apoptosis-induc-

Proteome-wide analysis of head and neck cancer ing factor (AIF) and Bag-1.31,32 Furthermore, overexpression of Hsp70 on the cell membrane of HNSCC cells was evaluated as a potential target for natural killer (NK) cells on tumor material and control tissues of HNSCC patients.33 On the other hand, Hsp90, which is elevated in HNSCC, may represent a novel target for cancer treatment, as 17allylamino-17-demethoxygeldanamycin (17-AAG), a benzoquinoid ansamycin antibiotic currently in Phase I clinical trial, has been recently shown that by binding to Hsp90 can cause the destabilization of various Hsp90-dependent kinases important in oncogenesis, and consequently can exert a desired anticancer effect.34,35 For example, Hsp90 forms a complex with Akt, and the use of 17-AAG has been demonstrated to reduce not only the expression levels of the kinase but also its intrinsic activity, which is elevated in HNSCC, thus providing a rational for evaluating 17-AAG in this tumor type.36,37 Proteins usually expressed at very low level and generally difficult to detect, were also identified in our analysis. These include Wnt-6, Wnt-14 and, placental growth factor (PIGF). The fact that these proteins could be detected at all in such limited starting material supports the sensitivity of our experimental approach, and the likely high levels of these molecules in normal or tumor tissues. In the case of the Wnt proteins, previous studies using gene arrays, including ours, have implicated the Wnt pathway in HNSCC pathogenesis.19,38 More recently, evidence was obtained that enhanced Wnt expression and signaling accelerate the progression of carcinomas by activating the epithelial–mesenchymal transition and local invasiveness.39 Nonetheless, very limited information is currently available specifically about Wnt-6 and Wnt-14. In regard to the Wnt-6, its expression in normal oral epithelium matches its pan-epithelial expression in normal epithelium in chicken, but it is also expressed at certain levels in cervical and colon cancer cells.40,41 Thus, further work will be required to study whether Wnt-6 is indeed expressed primarily in normal versus tumor squamous epithelium. Wnt14 is expressed in joint-forming regions of the skeleton during development, and its ectopic expression has been documented at the mRNA level in a variety of cancer cells, including those derived from brain, breast, pancreas, esophageal, gastric and cervical tumors.42,43 Thus, our findings suggest that Wnt-14 protein may correlate or even play a causative in carcinogenesis in HNSCC, an exciting possibility that warrants further investigation. PIGF was also detected in tumor samples and its potent pro-angiogenic function, through the activation of

197 the flt-1 vascular endothelial growth factor (VEGF) receptor (VEGFR-1) has been well documented, but very limited information is available on its putative role in tumor-induced angiogenesis. Nonetheless, available information suggests that PIGF is expressed in a variety of tumors, including meningiomas, prostate cancer, and renal cell cancer.44,45 Of more direct relevance to our present study, high expression levels of PIGF have been recently documented in mouse models of skin SCC.46 Thus, these observations raise the exciting possibility that this poorly studied angiogenic factor may play an unexpected role in the process of neovascularization that characterizes both experimental as well as human squamous carcinogenesis. In summary, we provide evidence that the use of recently developed LC–MS/MS techniques combined with laser-assisted microdissection and capture of human cells from clinical samples may enable a proteome-wide analysis of proteins expressed in normal and cancerous tissues. The emerging information is expected to provide valuable information regarding the still unknown molecular mechanisms promoting the malignant conversion of the normal oral epithelium, as well as may help identify biomarkers of disease development and progression, and novel therapeutic targets for HNSCC.

References 1. Jemal A, Tiwari RC, Murray T et al. Cancer statistics. CA Cancer J Clin 2004;54(1):8–29. 2. Shiu MN, Chen TH. Impact of betel quid, tobacco and alcohol on three-stage disease natural history of oral leukoplakia and cancer: implication for prevention of oral cancer. Eur J Cancer Prev 2004;13(1):39–45. 3. Ha PK, Benoit NE, Yochem R et al. A transcriptional progression model for head and neck cancer. Clin Cancer Res 2003;9(8):3058–64. 4. Bettendorf O, Piffko J, Bankfalvi A. Prognostic and predictive factors in oral squamous cell cancer: Important tools for planning individual therapy? Oral Oncol. 2004;40(2):110–9. 5. Neville BW, Day TA. Oral cancer and precancerous lesions. CA Cancer J Clin 2002;52(4):195–215. 6. Bankfalvi A, Piffko J. Prognostic and predictive factors in oral cancer: The role of the invasive tumour front. J Oral Pathol Med 2000;29(7):291–8. 7. Alevizos I, Mahadevappa M, Zhang X et al. Oral cancer in vivo gene expression profiling assisted by laser capture microdissection and microarray analysis. Oncogene 2001;20(43):6196–204. 8. Strausberg RL, Greenhut SF, Grouse LH, Schaefer CF, Buetow KH. In silico analysis of cancer through the Cancer Genome Anatomy Project. Trends Cell Biol 2001;11(11): S66–71.

198 9. Leethanakul C, Knezevic V, Patel V et al. Gene discovery in oral squamous cell carcinoma through the Head and Neck Cancer Genome Anatomy Project: confirmation by microarray analysis. Oral Oncol 2003;39(3):248–58. 10. Wasinger VC, Corthals GL. Proteomic tools for biomedicine. J Chromatogr B Anal Technol Biomed Life Sci 2002; 771(1–2):33–48. 11. Ong SE, Pandey A. An evaluation of the use of twodimensional gel electrophoresis in proteomics. Biomol Eng 2001;18(5):195–205. 12. Sickmann A, Mreyen M, Meyer HE. Mass spectrometry–a key technology in proteome research. Adv Biochem Eng Biotechnol 2003;83:141–76. 13. Bodnar WM, Blackburn RK, Krise JM, Moseley MA. Exploiting the complementary nature of LC/MALDI/MS/MS and LC/ESI/ MS/MS for increased proteome coverage. J Am Soc Mass Spectrom 2003;14(9):971–9. 14. Simone NL, Bonner RF, Gillespie JW, Emmert-Buck MR, Liotta LA. Laser-capture microdissection: opening the microscopic frontier to molecular analysis. Trends Genet 1998;14(7):272–6. 15. Wu SL, Hancock WS, Goodrich GG, Kunitake ST. An approach to the proteomic analysis of a breast cancer cell line (SKBR-3). Proteomics 2003;3(6):1037–46. 16. Zhang S, Van Pelt CK, Henion JD. Automated chip-based nanoelectrospray–mass spectrometry for rapid identification of proteins separated by two-dimensional gel electrophoresis. Electrophoresis 2003;24(21):3620–32. 17. Anderson DC, Li W, Payan DG, Noble WS. A new algorithm for the evaluation of shotgun peptide sequencing in proteomics: support vector machine classification of peptide MS/MS spectra and SEQUEST scores. J Proteome Res 2003;2(2):137–46. 18. Lamaroon A, Pattanaporn K, Pongsiriwet S et al. Analysis of 587 cases of oral squamous cell carcinoma in northern Thailand with a focus on young people. Int J Oral Maxillofac Surg 2004;33(1):84–8. 19. Leethanakul C, Patel V, Gillespie J et al. Distinct pattern of expression of differentiation and growth-related genes in squamous cell carcinomas of the head and neck revealed by the use of laser capture microdissection and cDNA arrays. Oncogene 2000;19(28):3220–4. 20. Wu SL, Amato H, Biringer R, Choudhary G, Shieh P, Hancock WS. Targeted proteomics of low-level proteins in human plasma by LC/MSn: using human growth hormone as a model system. J Proteome Res 2002;1(5):459–65. 21. Chaurand P, Schwartz SA, Caprioli RM. Assessing protein patterns in disease using imaging mass spectrometry. J Proteome Res 2004;3(2):245–52. 22. Peng J, Gygi SP. Proteomics: the move to mixtures. J Mass Spectrom 2001;36(10):1083–91. 23. Glanzer JG, Eberwine JH. Expression profiling of small cellular samples in cancer: less is more. Br J Cancer 2004;90(6):1111–4. 24. Li Zang DPT, Hancock WS, Sgroi DC, Karger BL. Proteomic analysis of ductal carcinoma of the breast using laser capture microdissection, LC–MS, and 16O/18O isotopic labeling. J Proteome Res 2004;3(3):604–12. 25. Real FX, Vila MR, Skoudy A, Ramaekers FC, Corominas JM. Intermediate filaments as differentiation markers of exocrine pancreas. II. Expression of cytokeratins of complex and stratified epithelia in normal pancreas and in pancreas cancer. Int J Cancer 1993;54(5):720–7. 26. Morifuji M, Taniguchi S, Sakai H, Nakabeppu Y, Ohishi M. Differential expression of cytokeratin after orthotopic

H. Baker et al.

27.

28.

29.

30.

31.

32.

33.

34. 35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

implantation of newly established human tongue cancer cell lines of defined metastatic ability. Am J Pathol 2000;156(4):1317–26. Prasad S, Thraves P, Dritschilo A, Kuettel M. Expression of cytokeratin-19 as a marker of neoplastic progression of human prostate epithelial cells. Prostate 1998;35(3): 203–11. Leunig A, Betz CS, Mehlmann M et al. Detection of squamous cell carcinoma of the oral cavity by imaging 5aminolevulinic acid-induced protoporphyrin IX fluorescence. Laryngoscope 2000;110(1):78–83. He QY, Chen J, Kung HF, Yuen AP, Chiu JF. Identification of tumor-associated proteins in oral tongue squamous cell carcinoma by proteomics. Proteomics 2004;4(1):271–8. Kaur J, Srivastava A, Ralhan R. Expression of 70 kDa heat shock protein in oral lesions: marker of biological stress or pathogenicity. Oral Oncol 1998;34(6):496–501. Schmitt E, Parcellier A, Gurbuxani S et al. Chemosensitization by a non-apoptogenic heat shock protein 70-binding apoptosis-inducing factor mutant. Cancer Res 2003;63(23):8233–40. Townsend PA, Cutress RI, Sharp A, Brimmell M, Packham G. BAG-1: a multifunctional regulator of cell growth and survival. Biochim Biophys Acta 2003;1603(2):83–98. Kleinjung T, Arndt O, Feldmann HJ et al. Heat shock protein 70 (Hsp70) membrane expression on head-and-neck cancer biopsy-a target for natural killer (NK) cells. Int J Radiat Oncol Biol Phys 2003;57(3):820–6. Neckers L, Ivy SP. Heat shock protein 90. Curr Opin Oncol 2003;15(6):419–24. Neckers L. Development of small molecule Hsp90 inhibitors: utilizing both forward and reverse chemical genomics for drug identification. Curr Med Chem 2003;10(9):733–9. Solit DB, Basso AD, Olshen AB, Scher HI, Rosen N. Inhibition of heat shock protein 90 function down-regulates Akt kinase and sensitizes tumors to Taxol. Cancer Res 2003;63(9):2139–44. Amornphimoltham P, Sriuranpong V, Patel V, Benavides F, Conti CJ, Sauk J et al. Persistent activation of the Akt pathway in head and neck squamous cell carcinoma: a potential target for UCN-01. Clin Cancer Res 2004;10(12): 4029–37. Rhee CS, Sen M, Lu D et al. Wnt and frizzled receptors as potential targets for immunotherapy in head and neck squamous cell carcinomas. Oncogene 2002;21(43): 6598–605. Uraguchi M, Morikawa M, Shirakawa M, Sanada K, Imai K. Activation of WNT family expression and signaling in squamous cell carcinomas of the oral cavity. J Dent Res 2004;83(4):327–32. Schubert FR, Mootoosamy RC, Walters EH et al. Wnt6 marks sites of epithelial transformations in the chick embryo. Mech Dev 2002;114(1–2):143–8. Kirikoshi H, Sekihara H, Katoh M. WNT10A and WNT6, clustered in human chromosome 2q35 region with head-totail manner, are strongly coexpressed in SW480 cells. Biochem Biophys Res Commun 2001;283(4):798–805. Hartmann C, Tabin CJ. Wnt-14 plays a pivotal role in inducing synovial joint formation in the developing appendicular skeleton. Cell 2001;104(3):341–51. Kirikoshi H, Sekihara H, Katoh M. Molecular cloning and characterization of WNT14B, a novel member of the WNT gene family. Int J Oncol 2001;19(5):947–52. Autiero M, Luttun A, Tjwa M, Carmeliet P. Placental growth factor and its receptor, vascular endothelial growth factor

Proteome-wide analysis of head and neck cancer receptor-1: novel targets for stimulation of ischemic tissue revascularization and inhibition of angiogenic and inflammatory disorders. J Thromb Haemost 2003;1(7): 1356– 70. 45. Matsumoto K, Suzuki K, Koike H et al. Prognostic significance of plasma placental growth factor levels in renal cell cancer: an association with clinical characteristics and

199 vascular endothelial growth factor levels. Anticancer Res 2003;23(6D):4953–8. 46. Larcher F, Franco M, Bolontrade M et al. Modulation of the angiogenesis response through Ha-ras control, placenta growth factor, and angioprotein expression in mouse skin carcinogenesis. Mol Carcinog 2003;37(2):83– 90.