Identification of PRDX4 and P4HA2 as Metastasis-Associated Proteins

ARTICLE pubs.acs.org/jpr

Identification of PRDX4 and P4HA2 as Metastasis-Associated Proteins in Oral Cavity Squamous Cell Carcinoma by Comparative Tissue Proteomics of Microdissected Specimens Using iTRAQ Technology Kai-Ping Chang,† Jau-Song Yu,‡,§ Kun-Yi Chien,‡,§ Chien-Wei Lee,§ Ying Liang,‡ Chun-Ta Liao,† Tzu-Chen Yen,|| Li-Yu Lee,^ Ling-Ling Huang,‡ Shiau-Chin Liu,† Yu-Sun Chang,‡,§ and Lang-Ming Chi‡,z,* †

Departments of Otolaryngology-Head & Neck Surgery, Chang Gung Memorial Hospital, Tao-Yuan, Taiwan Molecular Medicine Research Center, Chang Gung University, Tao-Yuan, Taiwan § Graduate Institute of Biomedical Sciences, College of Medicine, Chang Gung University, Tao-Yuan, Taiwan Department of Nuclear Medicine, Chang Gung Memorial Hospital, Tao-Yuan, Taiwan ^ Department of Pathology, Chang Gung Memorial Hospital, Tao-Yuan, Taiwan z Department of Medical Research and Development, Chang Gung Memorial Hospital, Tao-Yuan, Taiwan

)

‡

bS Supporting Information ABSTRACT: Cervical lymph node metastasis represents the major prognosticator for oral cavity squamous cell carcinoma (OSCC). Here, we used an iTRAQ-based quantitative proteomic approach to identify proteins that are differentially expressed between microdissected primary and metastatic OSCC tumors. The selected candidates were examined in tissue sections via immunohistochemistry, and their roles in OSCC cell function investigated using RNA interference. Seventyfour differentially expressed proteins in nodal metastases, including PRDX4 and P4HA2, were identified. Immunohistochemical analysis revealed significantly higher levels of PRDX4 and P4HA2 in tumor cells than adjacent non-tumor epithelia (P < 0.0001 and P < 0.0001, respectively), and even higher expression in the 31 metastatic tumors of lymph nodes, compared to the corresponding primary tumors (P = 0.060 and P = 0.002, respectively). Overexpression of PRDX4 and P4HA2 was significantly associated with positive pN status (P = 0.048 and P = 0.021, respectively). PRDX4 overexpression was a significant prognostic factor for disease-specific survival in both univariate and multivariate analyses (P = 0.034 and P = 0.032, respectively). Additionally, cell migration and invasiveness were attenuated in OEC-M1 cells upon in vitro knockdown of PRDX4 and P4HA2 with specific interfering RNA. Novel metastasisrelated prognostic markers for OSCC could be identified by our approach. KEYWORDS: oral cancer, OSCC, metastasis, PRDX4, P4HA2, iTRAQ-based proteomics, LCM

’ INTRODUCTION Oral cavity squamous cell carcinoma (OSCC) is the most common head-and-neck cancer, which accounts for approximately 3% of all newly diagnosed cancer cases.1,2 Despite recent advances in surgical, radiotherapy, and chemotherapy treatment protocols, the long-term survival of OSCC patients has remained at ∼50
60% over the past three decades.3,4 These unsatisfactory results and lack of improvement despite treatment modality advancements can be mainly explained by the fact that OSCCs are associated with high probability of cervical lymph node metastasis, historically regarded as the major poor prognosticator for OSCC treatment.5
8 Due to the significant impact of cervical metastasis on patient survival, in the current management of OSCC, clinically node-negative (cN0) and cN+ patients usually undergo elective and comprehensive neck dissection, respectively, involving surgical removal of cervical lymph nodes, followed by r 2011 American Chemical Society

pathological examination and staging (pN status). Although elective neck dissection for cN0 patients contributes to better locoregional control and survival,9,10 only one-third of the patients have true pN+ disease.7,11 On the other hand, approximately 15% of cN+ patients are truly node-negative, based on histopathological examination.7,10 These findings indicate that many patients are, in fact, overtreated with neck surgery. Therefore, better prediction of cervical metastasis with modern molecular methods may reduce patient exposure to unnecessary surgical procedures. Cervical metastasis is a major problem and prognosticator for OSCC management, and consequently, understanding its molecular basis is of significant importance. Currently, modern highthroughput molecular technology platforms are widely used to Received: April 3, 2011 Published: August 22, 2011 4935

dx.doi.org/10.1021/pr200311p | J. Proteome Res. 2011, 10, 4935–4947

Journal of Proteome Research

ARTICLE

Figure 1. Schematic representation of the experimental design, bioinformatics analysis and candidate validation for comparative proteome discovered in microdissected OSCC tissue specimens using iTRAQ labeling technology in this report.

identify the OSCC metastasis-related tumor markers. Previous cDNA microarray studies focusing on mRNA expression analysis of primary tumors have established the genetic expression profiles associated with OSCC lymphatic metastasis.11
15 However, association studies based on a systematic proteomic approach to reveal alterations in protein abundance are lacking. The recent development of state-of-the-art proteomic technologies has provided an opportunity to directly access the abundance and distribution of proteins in various clinical sample types.16,17 Additionally, the successful use of laser capture microdissection (LCM) for the isolation of a pure tumor cell population from heterogeneous cell types in tumor tissues and its integration with the proteomic approach has facilitated analysis of the protein profiles of particular cell types (e.g., metastasizing tumor cells) from tumor tissues.18
22 Previously, we established the proteome profiles derived from the OSCC cell secretome and compared these profiles between primary tumors and adjacent non-tumor epithelia from OSCC patients.19,23 In the present study, we have adapted a novel approach to identify metastasis-associated proteins in metastatic OSCC tumor cells. Considering the heterogeneity of cell population in tissues, laser capture microdissection was employed to capture tumor cells from metastatic cervical lymph nodes and the corresponding primary tumor in the oral cavity to improve the chance of tumor cell-related protein identifications. Proteins were extracted from the samples and labeled with iTRAQ tags (isobaric tags for relative and absolute quantitation), followed by liquid chromatography-tandem mass spectrometry analysis. From the differentially expressed protein database generated using this

approach, we identified 60 and 14 proteins displaying increased and decreased levels, respectively, in metastatic OSCC cells. The patient-to-patient fluctuation may cause a biased representation of the quantification results, but there is still no evidence-based standard for the amount of samples that must be used to correct the sample heterogeneity. Here, we consider the potential biomarkers discovered from limit amount of samples by quantitative proteomics should be verified in another large cohort by antibody-based validation to see whether it has a chance to be a good marker in future. Two novel proteins associated with OSCC metastasis, peroxiredoxin 4 (PRDX4) and prolyl 4-hydroxylase, alpha polypeptide II (P4HA2), involved in perioxidase activity and oxidation reduction of the enriched GO terms were selected for further validation in a large retrospective cohort of patients (Figure 1). Immunohistochemical staining revealed a significant association between PRDX4 and P4HA2 overexpression in primary tumor cells and positive pN status. Moreover, PRDX4 overexpression was a significant prognostic factor in both univariate and multivariate survival analyses. To our knowledge, this is the first study to use this novel approach of integrating LCM for sample preparation and iTRAQ labeling for comparative proteome analysis to identify potential markers of OSCC metastasis.

’ MATERIALS AND METHODS Patient Characteristics and Clinical Specimens

Three pairs of surgically resected primary and metastatic nodal tumor specimens obtained from three OSCC patients were used for LCM, as described previously.19 These samples were further 4936

dx.doi.org/10.1021/pr200311p |J. Proteome Res. 2011, 10, 4935–4947

Journal of Proteome Research subjected to comparative tissue proteome analysis using the iTRAQ technique. Tumor specimens for immunohistochemical analysis were obtained from 88 OSCC patients diagnosed at the Chang Gung Memorial Hospital (Tao-Yuan, Taiwan) from 2003 to 2006. Patients with inoperable disease, synchronous cancers, recurrent cancers, distant metastasis, or preoperative history of another malignancy were excluded from analysis. All patients provided informed consent prior to study participation, which was approved by the Institutional Review Board. The characteristics of the participating patients are presented in Supporting Information (Table S1). Patients underwent standard preoperative work-up according to institutional guidelines, including detailed medical history, complete physical examination, computed tomography or magnetic resonance imaging scans of the head and neck, chest radiographs, bone scan, and abdominal ultrasound. Primary tumors were intraoperatively excised with adequate margins under frozen section control. Classic radical or modified neck dissection (levels I
V) was performed in patients with clinically positive lymph node disease. Supraomohyoid neck dissection (levels I
III) was performed in clinically nodenegative patients.24 Surgical defects were immediately reconstructed via free flap or local flap by plastic surgeons where necessary. After surgical treatment, the pathological and nodal stages of all tumors were established according to the AJCC Cancer Staging Manual (2002). Post-operative radiotherapy was performed on patients with pathologic T4 tumors and positive lymph nodes within 6 weeks following surgery. Patients with pathologic multiple neck lymph nodes metastasis and/or extracapsular spread received adjuvant concurrent chemoradiotherapy. The chemotherapy regimen was cisplatin-based, involving a total radiation dose of 66 Gy. The prescribed dose was 1.8
2 Gy per day, 5 days per week. After discharge, all patients had regular follow-up visits every 2 months for the first year, every 3 months for the second year, and every 6 months thereafter.25,26 iTRAQ Labeling and Fractionation of Labeled Peptides

Proteins were extracted from microdissected cells using 0.1% RapiGest SF (Waters, Corporation, Milford, MA) in 50 mM Tris buffer, pH 8.0, and digested with sequencing grade modified trypsin (Promega, Madison, WI) at 37 C overnight. The desalted peptides were labeled with iTRAQ reagents (Applied Biosystems, Foster City, CA), according to the manufacturer’s instructions. All labeled peptides (100 μg) were mixed and separated on a 2  100 mm Gemini C18 column (Phenomenex, Torrance, CA) coupled to a Breeze HPLC instrument (Waters Corporation, Milford, MA) using an acetonitrile gradient with 0.05% ammonium hydroxide. Effluents were collected and pooled into about 40 fractions, depending on signal intensity at OD220/280. Each fraction was dried in vacuum and redissolved in 0.5% formic acid solution for subsequent LC
MS/MS analysis. MS-based Protein Identification and Quantification

Peptide samples were analyzed on a nanoLC-LTQ-Orbitrap hybrid mass spectrometer (Thermo Fisher, San Jose), as described previously.27 Four MS/MS microscans were averaged to create proper signal intensity for peptide quantification using iTRAQ reporter ions. The MS and MS/MS raw data were converted to MGF peak list files with DTASuperCharge 1.37 using SmartPicking with 200 segment size (Th) and max search level of 8. Generated MGF files were searched against a SwissProt human protein database (v.56; 20333 sequences) using the MASCOT search engine (v.2.2, Matrix Science, London). Search parameters were defined as digestion with trypsin, fixed modification of

ARTICLE

Figure 2. Protein identification and quantification in two sets of proteomics analyses using iTRAQ technique. The protein identification matched ProteinProphet probability higher than 0.95 and with at least two unique peptides was used. The protein quantification was performed using Libra module used in the TPP platform.

methylthiolation at cysteine, variable modification of N-terminal acetylation of protein, conversion of N-terminal glutamine to pyro-glutamate, oxidation of methionine, and iTRAQ labeling at lysine residue and N-terminus of peptides. One missed cleavage, mass accuracy of 6 ppm on the parent ion, and 0.6 Da on fragment ions were allowed. Identified proteins were subjected to further filtering using open-source proteomic statistical analysis software (Trans-Proteomic Pipeline, TPP v4.3 JETSTREAM rev 1, http://www.proteomecenter.org/software.php) developed at the Institute for Systems Biology (Seattle, WA), with PeptideProphet and ProteinProphet probability cutoffs set at 0.95. All protein identification datasets were fit to an estimated false positive error rate of mean ( 3SD, covering more than 99% of the proteins) were considered as no change in this test. Therefore, 1.5 fold-change was used as cutoff value for selecting upregulated and downregulated candidates in this study. Proteins whose LN/PT ratios were higher than 1.5 or lower than 0.67 and which displayed consistent changes in all data sets were considered metastasis-associated candidates. Consequently, 60 up-regulated and 14 down-regulated proteins in OSCC nodal metastasis were identified (listed in Table 1).

Among these proteins, 12 have been validated as OSCC metastasis markers in previous studies (Table 1).31
45 A further 9 proteins have been reported as dysregulated in metastasis of other cancer types (Table 1).46
57 To our knowledge, the remaining 53 proteins have not yet been identified as cancer metastasisrelated (Table 1). Among the 12 proteins known associated with OSCC metastasis in our candidate list, eleven and one proteins were upregulated and down-regulated in metastatic tumor cells, respectively. These results were consistent with many previous OSCC studies. In brief, Annexin A5 (ANXA5), one of the proteins up-regulated in metastatic OSCC tumors, belongs to the group of annexins involved in tumor development and progression, and affects the migration and invasion of OSCC cell lines.31 Fibronectin 1 (FN1) regulates cell adhesion and migration by modulating 4940

dx.doi.org/10.1021/pr200311p |J. Proteome Res. 2011, 10, 4935–4947

Journal of Proteome Research

ARTICLE

Figure 3. Overexpression of PRDX4 and P4HA2 in OSCC cells. (A) Immunohistochemical staining of PRDX4 and P4HA2 in the pericancerous adjacent normal epithelia (NE), tumor tissues and metastatic lymph nodes (LN) from two representative cases (scale bar = 100 μm). Expression patterns (brown staining) of PRDX4 and P4HA2 indicated that the proteins are localized in the cytoplasm of tumor cells. (B) Statistical analyses of immunohistochemical scores of PRDX4 and P4HA2 detected in the 74 paired samples revealed higher levels in tumor cells than nontumor normal epithelia (PRDX4: 168.4 ( 48.9 vs 18.9 ( 39.4, P < 0.0001; P4HA2: 160.1 ( 45.4 vs 8.3 ( 25.8, P < 0.0001). Furthermore, expression levels of both PRDX4 and P4HA2 were higher in the 31 metastatic tumors of lymph nodes than those in the corresponding primary tumors (PRDX4: 196.1 ( 42.1 vs 168.4 ( 48.9, P = 0.060; P4HA2: 201.9 ( 29.1 vs 160.1 ( 45.4, P = 0.002).

fibrillar networks of metastatic outgrowth.32,33 Integrin beta-6 (ITGB6) increased OSCC cells migration and invasion in vitro and in vivo.32 MMPs, including MMP10, MMP13 and MMP14, have additionally been implicated in tumor metastasis.34
38 Periostin (POSTN) is widely expressed in many malignant tumors, and related to positive nodal metastasis and poorer prognosis in OSCC.39,40 Secreted protein acidic and rich in cysteine (SPARC) is highly expressed in tongue squamous cell carcinoma, and positively correlated to survival, occult lymph node metastasis and recurrence.41 Thrombospondin-1 (THBS1), synthesized and accumulated in the stroma of OSCC tumors, has been demonstrated to promote the tumorigenesis by enhancing the motility and proteolytic activity in a paracrine manner.42 Tenascin-C (TNC) is an adhesion modulatory extracellular matrix molecule that induces tumor proliferation, angiogenesis, and metastasis and reduces survival prognosis in some cancers.43 Vimentin (VIM), a protein of the intermediate filament family, functions as an organizer of a number

of critical proteins involved in OSCC cell migration.44 On the other hand, most of down-regulated proteins in this study were newly discovered and have not been validated previously. Psoriasin (S100A7) is the only one that was confirmed to augment in early OSCC specimen but abate in cervical lymph node metastasis by quantitative real-time RT-PCR.45 Overall, these previous reports further confirm the validity of our approach and findings. Functional Significance of the Up-regulated Protein Candidates

To explore the biological significances of protein candidates in depth, the pathway annotations and functional relationship of candidate genes were analyzed using DAVID Bioinformatics. Pathway enrichment analysis revealed that the up-regulated proteins are significantly involved in the extracellular matrix (ECM)-receptor interaction, focal adhesion and integrin signaling pathways (all p < 0.001) (Table 2), highlighting the importance of these three pathways on the mechanisms of OSCC 4941

dx.doi.org/10.1021/pr200311p |J. Proteome Res. 2011, 10, 4935–4947

Journal of Proteome Research

ARTICLE

Table 4. Clinicopathological Characteristics Related to the Expression of PRDX4 and P4HA2 in OSCC Specimens PRDX4 overexp (%) no

yes

P4HA2 overexp (%) p-value

no

29.5
85.0

30.1
83.9

49.9 ( 13.2

50.6 ( 11.5

0.589 0.064

yes

p-value

Age range

29.5
82.5

30.1
85.0

mean ( SD

47.4 ( 11.5

52.4 ( 12.1

0.057

T1, 2

19 (44)

24 (56)

0.852

T3, 4

19 (42)

26 (58)

29 (51) 9 (29)

28 (49) 22 (71)

0.048a

I
II

15 (52)

14 (48)

0.256

III
IV

23 (39)

36 (61)

W-D + M-D

35 (43)

47 (57)

P-D

3 (50)

3 (50)

26 (43)

35 (57)

12 (44)

15 (56)

pT status 11 (26)

32 (74)

20 (44)

25 (56)

25 (44) 6 (19)

32 (6) 25 (81)

0.021a

10 (34)

19 (66)

0.918

21 (36)

38 (64)

pN status N(
) N(+) Overall stage

Differentiationb

Perineural invasion negative positive

0.726

0.873

29 (35)

53 (65)

2 (33)

4 (67)

26 (41)

36 (58)

5 (19)

21 (81)

1.000

0.041a

Tumor depth (mm) range

1.0
48.0

0.5
40.0

mean ( SD

11.8 ( 9.9

11.2 ( 8.2

0.863

1.5
48.0

0.5
40.0

13.3 ( 11.4

10.8 ( 7.6

0.660

13 (35)

24 (65)

0.194

25 (59)

26 (51)

12 (33)

24 (67)

0.757

19 (37)

33 (63)

no

4 (31)

9 (69)

yes

34 (45)

41 (55)

no

5 (45)

6 (55)

2 (18)

9 (64)

yes

33 (43)

44 (57)

29 (38)

48 (62)

38

50

31

57

Alcohol consumption no yes Betel nut chewing

0.327

5 (38)

8 (62)

26 (35)

49 (65)

0.764

Smoking

Total case number

0.870

0.205

a

Statistically significant; overexp: overexpression. b W-D: well-differentiated squamous cell carcinoma; M-D: moderatively differentated squamous cell carcinoma; P-D: poorly differentiated squamous cell carcinoma.

metastasis. The significance of these pathways has been previously confirmed by the associated individual proteins analyzed in many studies.32,33,42,43,46,48,54 The enrichment GO terms are analyzed using DAVID Functional Classification Tool and listed in Table 3. Consistent with pathway analyses, the up-regulated genes grouped in the first two classes of the enriched GO terms are mainly extracellular proteins that have been demonstrated to promote tumor metastasis of many malignancies.15,32,33,39,40,43,46,48,53 These results collectively support the feasibility of our approach for identifying cancer metastasis-related proteins, and suggest that these proteins represent good candidate prognostic markers of OSCC nodal metastasis. To confirm data obtained from iTRAQ-based quantification, 4 novel up-regulated candidates with availability of good antibodies, PRDX4, P4HA2, calponin 3 (CNN3) and caldesmon 1 (CALD1), whose roles in OSCC have not been addressed before, were further verified by immunohistochemical study. CNN3, a member of calponin family originally identified as molecules binding to F-actin, calmodulin, and tropomyosin, is involved in regulating the contraction/relaxation cycle in smooth

muscle cells.58 CALD1 is a major actomyosin-binding protein containing the ability of binding to actin, tropomyosin, Ca(2+)calmodulin, myosin, and phospholipids and crucially involved in the remodeling of the microfilament network. Therefore, it is an important regulator of various cell functions, including cell motility.59,60 Nonetheless, the role of these two proteins in cancer metastasis and OSCC tumors has not been disclosed yet. PRDX4 is one of the peroxiredoxin (PRDX) family of proteins consisting of six identified mammalian isoforms that perform a dual function as small H2O2 scavengers in normal cells and a key survival factor in tumor cells.61 PRDXs are increasingly expressed in malignant cells, compared to benign cells, and protect tumor cells from oxidative stress-induced apoptosis in different cancer types.61,62 Although PRDX4 is overexpressed in different cancers, our current understanding of its specific function in cancer cells is limited. PRDX4 is associated with tumor metastasis in the colon at the mRNA level.52 P4HA2 is the subunit alpha-2 of collagen prolyl 4-hydroxylases (C
P4Hs), which catalyze the formation of 4-hydroxyproline residues essential for proper structural folding of newly synthesized proteins (e.g., procollagen) and play a pivotal 4942

dx.doi.org/10.1021/pr200311p |J. Proteome Res. 2011, 10, 4935–4947

Journal of Proteome Research

ARTICLE

Figure 4. Overexpression of PRDX4 is associated with poorer prognosis with regard to overall and disease-specific survival. (A) Kaplan
Meier plot shows that the 5-year overall survival for patient subgroups stratified by overexpression of PRDX4 and P4HA2 was 70.6 versus 46.3% (P = 0.061) and 63.0 versus 57.3% (P = 0.702), respectively. (B) Kaplan
Meier plot for disease-specific survival shows that the 5-year disease-specific survival for patient subgroups stratified by overexpression of PRDX4 and P4HA2 was 78.4 versus 47.3% (P = 0.035) and 68.8 versus 59.8% (P = 0.380), respectively.

Table 5. Cox Proportional Hazard Models on Disease-Specific Survival univariate

multivariate

crude HR (95% CI)

p-value

adjusted HR (95% CI)a

p-value

2.334 (1.048
5.199)

0.038b

2.774 (1.168
6.588)

0.020b

3.465 (1.603
7.492)

0.001b

3.022 (1.302
7.015)

0.010b

2.882 (0.864
9.620)

0.085

6.730 (1.565
28.932)

0.010b

Positive vs Negative

1.970 (0.910
4.262)

0.085

1.036 (0.414
2.592)

0.939

PRDX4 overexpression Yes vs No

2.546 (1.072
6.050)

0.034b

pT status T3-T4 vs T1-T2 pN status N(+) vs N(
) c

Differentiation

P-D vs W-D + M-D Perineural invasion

a

0.032b

2.774 (1.087
7.075) b

c

HR: Hazard Ratio; CI: Confidence Interval; multivariate analysis also adjusted with age and sex. Statistically significant. W-D: well-differentiated squamous cell carcinoma; M-D: moderatively differentated squamous cell carcinoma; P-D: poorly differentiated squamous cell carcinoma.

role in cell-matrix interactions.63 Vertebrate C
P4Hs are alpha2beta2 tetramers with three isoenzymes that share identical beta subunits but differ in their catalytic alpha subunits (P4HA1, P4HA2 and P4HA3 for C
P4H I, II, and III, respectively).64 All P4HAs are oxygen-sensitive and inducible in response to hypoxia and transforming growth factor-beta stimulation.65,66 In the present study, P4HA2 has been identified as a novel metastasis-associated protein. Collagen prolyl 4-hydroxylase subunit alpha-1 (P4HA1), another dominant alpha subunit of C
P4Hs, overexpressed in OSCC tumor cells in a previous study by our group,19 was slightly increased in nodal metastatic cells (Supporting Information, Table S2). Although the roles of P4HA2 in tumorigenesis are obscure at present, its overexpression has been reported in thyroid and breast cancers and P4HA2 is listed as one of the genes associated with cervical metastasis of head-and-neck squamous cell carcinomas in several genomic expression profiling studies.67
69 These results support the

potential significance of P4HA2 in carcinogenesis as well as metastatic mechanisms of cancer cells. Overexpression of PRDX4 and P4HA2 in Tumor Cells of OSCC Specimens

Tissue sections from 88 patients were subjected to immunohistochemical (IHC) staining. Both PRDX4 and P4HA2 were highly expressed in the cytoplasm of tumor cells, with no nuclear or membranous staining. Expression of both proteins was mainly absent or extremely low in tumor-infiltrating lymphocytes and interstitial tissues. Moreover, paired normal oral epithelium samples showed relatively low or no expression of PRDX4 and P4HA2 (Figure 3A). To ascertain whether the two proteins are specifically overexpressed in metastatic tumor cells, the corresponding resected metastatic lymph nodes were additionally subjected to IHC staining. Accordingly, both PRDX4 and P4HA2 were highly expressed in the cytoplasm of metastatic tumor cells, compared to adjacent tissue and nontumor cells in 4943

dx.doi.org/10.1021/pr200311p |J. Proteome Res. 2011, 10, 4935–4947

Journal of Proteome Research

ARTICLE

Figure 5. Attenuation of OSCC cell migration and invasiveness by PRDX4-specific and P4HA2-specific RNAi. (A) Western blot analysis for PRDX4 and P4HA2 in lysates from OEC-M1 cells transfected with PRDX4-specific RNAi (si-PRDX4), P4HA2-specific RNAi (si-P4HA2) or scrambled sequence control RNAi for 48 and 96 h. (B) Cell proliferation assay for OEC-M1 cells transfected with si-PRDX4, si-P4HA2 or scrambled sequence control RNAi (si-CTL) for 48 and 96 h. (C, D) Cell migration and invasiveness assays for OEC-M1 cells transfected with si-PRDX4, si-P4HA2 or si-CTL for 48 h.

the metastatic lymph nodes (Figure 3A). Statistical analysis of the 74 paired samples available from these 88 OSCC patients disclosed that the IHC scores of both PRDX4 and P4HA2 were significantly higher in tumor tissue than nontumor normal epithelium (PRDX4: 168.4 ( 48.9 vs 18.9 ( 39.4, P < 0.0001; P4HA2: 160.1 ( 45.4 vs 8.3 ( 25.8, P < 0.0001) (Figure 3B). Furthermore, PRDX4 and P4HA2 levels were markedly higher in the 31 metastatic tumors of lymph nodes, compared to those in the corresponding primary tumors (PRDX4: 196.1 ( 42.1 vs 168.4 ( 48.9, P = 0.060; P4HA2: 201.9 ( 29.1 vs 160.1 ( 45.4, P < 0.002) (Figure 3B). Immunohistochemical analyses confirmed that both proteins were more highly expressed in tumor cells of metastatic lymph nodes than primary tumor cells and almost absent in normal oral epithelia. Increased Expression of CNN3 and CALD1 in the Metastatic OSCC Cells

To validate more novel protein markers from the candidates discovered, we also selected CNN 3 and CALD1 to confirm the results of our iTRAQ analyses. By immunohistochemistry, we found that both proteins were overexpressed in the tumor cells of the metastatic lymph nodes and the corresponding primary tumors from 31 patients (Figure S2and Figure S3, Supporting Information). Furthermore, expression levels of CNN3 were higher in the metastatic cancer cells of lymph nodes than those in

the corresponding primary tumors (165.9 ( 49.1 vs 140.8 ( 50.3, P = 0.004; Figure S2). Similarly, increased CALD1 expression was also found in the metastatic cancer cells of lymph nodes than those in the corresponding primary tumors (171.8 ( 35.3 vs 143.9 ( 47.4, P = 0.027; Figure S3). Association of PRDX4 and P4HA2 Expression with Various Clinicopathological Manifestations

Next, we evaluated the relationships between increased PRDX4 or P4HA2 expression and various clinicopathological characteristics of OSCC patients (Table 4). PRDX4 overexpression was significantly associated with positive pN status (P = 0.048), and P4HA2 with both positive pN status and perineural invasion (P = 0.021 and P = 0.041, respectively). However, we observed no association between PRDX4 or P4HA2 overexpression in OSCC tumors and patient age, pT status, overall pathological stages, cell differentiation, tumor depth, and habitual behaviors of betel nut chewing, cigarette smoking and alcohol consumption. Consistent with the hypothesis of our study approach, overexpression of both proteins was significantly associated with nodal metastasis (pN status). PRDX4 or P4HA2 Overexpression and Patient Overall Survival (OS)

On the basis of data from 88 patients undergoing complete standardized treatment and regular follow-up, we further evaluated 4944

dx.doi.org/10.1021/pr200311p |J. Proteome Res. 2011, 10, 4935–4947

Journal of Proteome Research whether overexpression of PRDX4 or P4HA2 was associated with patient OS. Survival analysis showed that the five-year OS rates for patient subgroups stratified by the absence or presence of PRDX4 were 70.6% and 46.3%, respectively. The mean survival time of the patients without and with PRDX4 overexpression in OS was 2.69 and 2.54 years, respectively. The hazard ratio of PRDX4 overexpression in OS was 2.066 (95% CI, 0.942
4.529; P = 0.070). These differences in OS were marginally significant when compared using the log-rank test (P = 0.061) (Figure 4A). Conversely, the 5-year OS rates established via Kaplan
Meier plots of patients stratified by the absence or presence of P4HA2 were not significantly different (63.0 and 57.3%, respectively; P = 0.702) (Figure 4A). The mean survival time of the patients without and with P4HA2 in OS was 3.05 and 2.77 years, respectively. The hazard ratio of P4HA2 overexpression in OS was 1.200 (95% CI, 0.559
2.574; P = 0.640). PRDX4 or P4HA2 Overexpression and Disease-Specific Survival (DSS)

In Kaplan
Meier plots, 5-year DSS rates for patients stratified by the absence or presence of PRDX4 overexpression were 78.4% and 47.3%, respectively. The mean survival time of the patients without and with PRDX4 overexpression in DSS was 1.58 and 2.74 years, respectively. The hazard ratio of PRDX4 overexpression in DSS was 2.546 (95% CI, 1.072
6.050; P = 0.034). These differences in DSS were statistically significant, as observed with the log-rank test (P = 0.035) (Figure 4B). On the other hand, the mean survival time of the patients without and with P4HA2 in DSS was 3.18 and 2.82 years, respectively. The hazard ratio of P4HA2 overexpression in DSS was 1.427 (95% CI, 0.622
3.275; P = 0.401). Five-year DSS rates for patients stratified by the absence or presence of P4HA2 was not significantly different with the log-rank test (68.8% and 59.8%, respectively; P = 0.380) (Figure 4B). Moreover, PRDX4 overexpression was a significant predictor of DSS in univariate analysis of the Cox proportional regression model (Table 5), but not P4HA2 overexpression. To determine whether PRDX4 overexpression can be applied as an independent predictor of DSS, multivariate analysis was performed using age, gender, pT status, pN status, overall stage, perineural invasion, tumor differentiation, and PRDX4 overexpression as parameters. Our results indicate that pT status, pN status, tumor differentiation, and PRDX4 overexpression are independent predictors of DSS (P = 0.020, 0.01, 0.010 and 0.032, respectively; Tables 5). Our data indicate that PRDX4 overexpression is critical in nodal metastatic processes and DSS in OSCC patients. PRDX4 and P4HA2 Promote OSCC Cell Migration and Invasion in vitro

To evaluate the biological significance of PRDX4 and P4HA2 in vitro, endogenous expression of these proteins in the OSCC cell line, OEC-M1, was eliminated using specific RNAi. The effects of RNAi were determined by Western blotting of OEC-M1 cells transfected with PRDX4-specific RNAi (si-PRDX4), P4HA2-specific RNAi (si-P4HA2) or a scrambled sequence control RNAi (si-CTL). As shown in Figure 5A, we observed a significant decrease in expression of endogenous PRDX4 (approximately 80% and 45% reduction at 48 and 96 h, respectively) and P4HA2 (at least 99% and 91% reduction at 48 and 96 h, respectively) in si-PRDX4and si-P4HA2-transfected cells, compared with control RNAi transfectants, respectively. Control, si-PRDX4- and si-P4HA2-transfected cells were subsequently analyzed for proliferation, migration, and invasiveness.

ARTICLE

Knockdown of PRDX4 or P4HA2 expression had a marginal effect (