A panel of biomarkers for skin squamous cell carcinoma: various functional entities and differential responses to resveratrol

Abstract

In contrast to the highly malignant melanoma, skin squamous cell carcinoma (SCC) usually presents with lower morbidity. However, its incidence has been alarmingly rising worldwide and is a public health burden, let alone the current SCC cancer classification scheme is inadequate. Due to its features of progressing along different pathologic stages, early detection of precancerous lesions with accurate molecular markers would be desirable for cancer prevention and treatment. In the present study, using immunohistochemical staining of 85 clinical samples, we profiled the expression of a panel of ten proteins from five functional divisions implicated in SCC development, i.e. cytokeratins, intercellular molecules, chaperone proteins, transcription factors, and mitochondrial redox enzymes. The differential alterations of the proteins in SCC cell lines SCL12 and COLO16, upon resveratrol therapy, were also examined by immunocytochemistry (ICC). Our data reveal that, while all these proteins show significant correlation with cancer initiation and/or progression, a comprehensive panel encompassing a range of biologic functions, instead of a single marker, will provide prognostic value in SCC diagnosis and management. Additionally, the strong correlation among the proteins with cancer stages implies their distinct roles in SCC pathogenesis and contributions to the therapeutic effects of resveratrol, which is demonstrated in the resveratrol-sensitive COLO16 cells, but not in the resveratrol-resistant SCL12 cells.

Introduction

Squamous cell carcinoma (SCC) is, by definition, an uncontrolled proliferation of squamous cells in the epidermal layer. SCC is the second most prevalent common malignant skin cancer next to basal cell carcinoma (BCC), making up 20% of all skin cancers [1]. As distinguished from melanoma-the most malignant type of skin cancer derived from the epidermis, SCC and BCC constitute a group of skin cancers termed non-melanocytic skin cancer (NMSCs) [2]. The risk of SCC increases in populations with Caucasian background and particularly with fair skin that is prone to sunburning rather than tanning, skin type I and II according to the Fitzpatrick scale [3]. Other predisposing factors for SCC incidence include long-term UV-light radiation and immune suppression. SCC frequently occurs in sun-exposed areas such as the face, neck, and dorsal hands [4]. In contrast to BCC, SCC is tightly associated with cumulative sun exposure and displays a histologically and clinically progressive continuum with precursor lesions including actinic keratosis (AK) and Bowen’s disease [5]. With an 8% recurrence and 5% metastasis rate within 5 years [6], it is more likely to invade surrounding dermis, nearby lymph nodes and distant organs than BCC [7]. With certain geographic variations, the incidence rates of NMSCs are on the rise alarmingly worldwide during the past decades and its increasing morbidity has become a tremendous economic burden calling for an improved registration system, accurate and early detection and effective management [1,8]. A positive association between NMSC and other forms of malignancy is also confirmed by a large body of cohort studies and meta-analysis of bibliographic databases [9-11].

Immunohistochemistry (IHC) has been the indispensable golden standard for skin cancer diagnosis and prognosis, using biomarkers of well-established panels or newly discovered molecules, such as pan cytokeratin [12] and micro RNA array [13]. IHC is also of significance in differential diagnosis between SCC, BCC, melanoma, Merkel cell carcinoma and other skin cancers bearing overlapping appearance [14,15]. However, a consensus has not yet been reached for reliable and distinctive biomarkers as signatures of SCC progression. In this work, we compared the protein profiles of noncancerous (normal), precancerous lesions, and SCC specimens, by mapping 10 selected proteins in groups of keratins (CK10, CK17), cell-cell adhesion factors (CD44, EZR, membrane E-cadherin and β-catenin), chaperone proteins (Hsp75 and Hsp90-α), transcription regulator EXOSC10 and mitochondrial redox protein SOD2. Given the progressive nature of SCC, tracing the footprint of the salient proteins that impart paramount impacts in disease outcome, will not only provide important clues to understanding cancer initiation and exacerbation, but also bring a new perspective to cancer prevention and treatment.

Resveratrol is a dietary polyphenol non-flavonoid compound, found in grapes, berries and peanuts. Its potential preventive and therapeutic effects in cancers, by affecting cellular events in cancer initiation, promotion and progression, have been widely documented [16-22]. While conventional treatment regimes pose many adverse responses, seeking well-tolerated effective anti-cancer agents, such as resveratrol, has become desirable [19,23]. Studies and clinical trials of resveratrol supplements in skin cancer, and the current situation and challenges are also discussed [24,25]. However, the majority of studies were focused on the viability of resveratrol alone or in combination with other standard treatment in melanoma, the most malignant but least common skin cancer. Research pertaining to its effects in SCC is extremely sparse and emphasizes protection against repetitive UV irradiation-induced cutaneous damage, including skin cancer, in humans and mice [26-28].

Apart from the bioavailability, metabolic regulation, and administration route, the uncertainty about the clinical use of resveratrol also resides in the variance of the individual molecular niche within the targets. It strongly prompts us to test the in situ expression of the cancer-stage related proteins in SCC cells displaying differential sensitivity to resveratrol, revealed by our previous work [29]. To the best of our knowledge, none of these proteins has been marked as a target of resveratrol in SCC treatment.

The evaluation of the correlation between novel biomarkers and cancer stage and drug responses may aid discovery of a new strategy to hamper or even to reverse cancer development and recurrence. In addition, the present study may strengthen the bridge between the protein profiling with proteomics and its therapeutic application.

Methods and materials

Selection of resection specimens

A total of 85 representative human epidermal cases were obtained from the tumour tissue bank established by the Cancer Centre of Dalian Medical University and the Key Laboratory of Cancer Genetics and Epigenetic, Liaoning Province, China. Specimens were parts of surgically removed tissue from patients admitted to the Dalian Dermatology Hospital between 1998 and 2010, without a history of chemo- or radiotherapy before the operation. Patients were aged 21-97 (mean 68.9), including 34 males (40%) and 51 females (60%). According to the clinicopathological diagnosis, specimens were divided into three groups: 42 normal tissue (non-cancerous regions surrounding the cancer), 34 precancerous lesions (3 Bowen’s disease, 24 actinic keratosis and 7 cutaneous horns) and 51 squamous cell cancer (SCC) tissues. The SCC specimens were further distinguished as well- (18), moderate- (13) and poorly- (20) differentiated subgroups.

Use of the tissues in this study was approved by the ethics committee of the Medical Faculty of Dalian Medical University, and samples were obtained with written informed consent of all patients or responsible family members.

Construction of paraffin embedded microarray

As described in our previous papers [30-32], a patented device (patent number: CN 02109826) owned by the laboratory was applied to produce high-density paraffin embedded microarray. Briefly, a paraffin mold block was firstly prepared and accurately positioned by coordinates. A matrix of punched holes (diameter of 0.6 mm) with a horizontal and vertical pitch of 1 mm was produced. The selected and labelled tissues were mounted onto the corresponding sites on the matrix by special sampler (internal diameter 0.6 mm). Serial sections of paraffin tissue at 4 μm thickness were used for hematoxylin-eosin (H/E) and immunohistochemical staining.

Cells and treatment

Human squamous cell carcinoma cell lines SCL12 (gift from Benjamin Franklin medical centre of Freie Universitat, Germany) and COLO16 (gift from the Institute for chemical physics of Chinese Academy of Sciences, China) were cultured in low glucose Dulbecco’s Modified Eagle’s Medium supplemented with 10% foetal bovine serum (FBS) and Penicillin-Streptomycin (all from Gibco) at 37°C in 5% CO₂. Resveratrol (Sigma, USA) was dissolved in dimethyl sulfoxide (DMSO) to make 100 mM stock solution and stored at -20°C. The working solution was made by diluting the stock with culture medium to 100 µM before use. Cells were treated for 48 hrs before being collected for different experimental purposes. Untreated cells served as controls. The experiments were repeated at least 3 times to confirm consistency.

Immunohistochemistry (IHC) and immunocytochemistry (ICC) examinations

The ten proteins of various functional implications in skin cancer development were selected for immunocytochemistry (ICC) on cell-bearing coverslips and tissue microarray-based immunohistochemical (IHC) on the paraffin sections of tissue microarray analyses, as described previously [33]. As described elsewhere [34], samples derived from surgical dissection were processed promptly through formaldehyde fixation, gradient dehydration, and paraffin embedding.

Antigen retrieval with citrate buffer (Sigma-Aldrich, USA) was performed for IHC examination. For both IHC and ICC procedures, 1% Triton X-100 was used to increase cell permeability, followed by endogenous peroxidase blocking by hydrogen peroxide (H₂O₂). Specimens were then incubated with the primary antibodies at 4°C overnight and the secondary antibodies at 37°C for 15 minutes, with thorough washing PBS in between. Sections were at last counterstained with hematoxylin. The intensity and distribution of the sections were reviewed independently by two experienced dermatopathologists to ensure unbiased judgement.

The monoclonal mouse anti-human antibodies against CK10, CK17, CD44, EZR, Hsp75, Hsp90-α, EXOSC10 and SOD2 were purchased from Santa Cruz Biotechnology, Inc. (USA), while the polyclonal rabbit anti-human E-cadherin and β-catenin antibodies were from Proteintech Group, Inc. (USA) and Bioworld Technology (USA), respectively. The sections incubated with the absence of the primary antibody were used as background control. Positive binding of the antibodies was labelled brown as a result of peroxidase-conjugated secondary antibodies (Envision⁺-system, DakoCytomation).

The staining outcome was determined by two parameters, each rigorously scored; i.e. ‘A’ for staining intensity and ‘B’ for the proportions of positively labelled cells out of the number of cells viewed. The value of ‘A’ was defined as follows, 0 for negative immune-labelling, 1 to 3 for samples slightly, moderately and strongly stained higher than the background, respectively. The value of ‘B’ for proportionality was defined as 0 for none of the cells were stained, 1 for < 1/3, 2 for 1/3~2/3 and 3 for ≥ 2/3 of positive labelling, respectively. The final staining results were calculated as the multiple value A*B, which indicated a rating index ranging from 0 to 9. While 0 meant negative (-), scores of 1-2, 3-4 and 6-9 were denoted ‘weakly positive (+)’, ‘moderate positive (++)’ and ‘strongly positive (+++)’, respectively. By this means, the intensity and distribution of the target protein were semi-quantified accurately.

Statistical analysis

SPSS version 17.0 software was used to analyse the experimental data. The non-parametric test Mann-Whitney was used to analyse the differences of indexes between various differentiation phases of SCC tissue, pre-cancerous lesions and normal tissue. Spearman Rank Correlation test was used to calculate the correlation between different protein expressions and disease progression; P ≤ 0.05 was considered statistically relevant.

Results

IHC revealed progressive signatures of the candidate biomarkers in tissues from normal to SCC

A total of 85 resection specimens were divided into 3 groups: normal (42), pre-cancerous lesions (34) and SCC (51), according to pathological evaluation. A more detailed classification was applied by distinguishing the pre-cancerous group as Bowen’s Disease (BD 3), actinic keratosis (24) and cutaneous horn (7), whereas the SCC samples were subdivided into well- (18), moderate- (13), and poorly- (20) differentiated subgroups.

Using immunohistochemistry, the samples were stained with antibodies against 10 proteins that functionally involve five distinct molecular entities: cytokeratins CK10 and CK17; cell adhesion-associated proteins CD44, EZR, membrane E-cadherin and β-catenin; chaperone proteins Hsp75 and Hsp90-α; transcription regulator EXOSC10; and mitochondrial redox protein SOD2.

As demonstrated in the representative staining images, the two cytokeratins showed opposite trends of change (Figure 1A). Whereas CK10 was lost in SCC tissues, CK17 expression was incrementally increased. Statistical analysis of protein expressions of CK10 and CK17 at different disease progress phases (Table 1) indicated significant differences between normal and SCC (P < 0.001), pre-cancerous, and SCC group (P < 0.001). The increase of CK17 expression from normal to pre-cancerous lesions was also significant (P < 0.001). While CK10 was 100% positive in normal tissues and slightly decreased to 94% at the pre-cancerous stage, it was reduced to 28% when cancer developed. In contrast, CK17 was positively stained in 9% of normal tissues, but gradually increased to 65% and 82% at the stages of pre-cancerous lesions and SCC respectively. CK10 and CK17 were strongly correlated with cancer development, inversely (rs = -0.626, P < 0.001) and positively (rs = 0.67, P < 0.001) respectively. There were no significant differences between differentiation stages within SCC. Further calculations indicated a significant inverse correlation between CK10 and CK17 expression (rs = -0.459, P < 0.001).

Figure 1.

Representative IHC staining of the candidate markers in groups of cytokerattins (CK10, CK17), cellular adhesion molecules (CD44, EZR), cell-cell junction proteins (membrane β-catenin, E-cadherin), chaperone proteins (Hsp90-α, Hsp75), RNA-binding protein (EXOSC10) and mitochondrial redox enzyme (SOD2) in normal, pre-cancerous, or SCC patient samples.

Table 1.

IHC staining of cytokeratins CK10/CK17 in clinical samples

			Intensity of CK10/CK17 staining
			-	+	++	+++
Normal	42	CK10	0 (0%)	0 (0%)	9 (21%)	33 (79%)	# ∨
		CK17	38 (91%)	1 (2%)	1 (2%)	2 (5%)	# ∧
Pre-cancerous	34	CK10	2 (6%)	0 (0%)	6 (18%)	26 (76%)	#
		CK17	12 (35%)	17 (50%)	3 (9%)	2 (6%)	#
Bowen’s disease	3	CK10	0 (0%)	0 (0%)	2 (67%)	1 (33%)
		CK17	3 (100%)	0 (0%)	0 (0%)	0 (0%)
Actinic keratosis	24	CK10	0 (0%)	0 (0%)	2 (8%)	22 (92%)
		CK17	9 (38%)	11 (46%)	3 (12%)	1 (4%)
Cutaneous horn	7	CK10	2 (29%)	0 (0%)	2 (29%)	3 (42%)
		CK17	0 (0%)	6 (86%)	0 (0%)	1 (14%)
SCC	51	CK10	37 (72%)	3 (6%)	0 (0%)	11 (22%)
		CK17	9 (18%)	3 (5%)	9 (18%)	30 (59%)
Well-differentiated	18	CK10	11 (61%)	1 (6%)	0 (0%)	6 (33%)
		CK17	0 (0%)	2 (11%)	5 (28%)	11 (61%)
Moderately-differentiated	13	CK10	10 (77%)	1 (8%)	0 (0%)	2 (15%)
		CK17	5 (38%)	1 (8%)	2 (16%)	5 (38%)
Poorly-differentiated	20	CK10	16 (80%)	1 (5%)	0 (0%)	3 (15%)
		CK17	4 (20%)	0 (0%)	2 (10%)	14 (70%)

#: Significantly different from SCC samples (P ≤ 0.05); ∨: Negatively correlated with cancer initiation (P ≤ 0.05); ∧: Positively correlated with cancer initiation (P ≤ 0.05).

Among the proteins involved in inter-cell adhesion, staining of both CD44 and EZR was progressively enhanced from normal to cancerous tissues (Figure 1B). There was a highly significant difference between normal and SCC group (17% to 61% in CD44, 21% to 96% in EZR, P < 0.001), pre-cancerous lesions and SCC group (30% to 61% in CD44, 44% to 96% in EZR, P < 0.001). Positive correlations were calculated as rs = 0.383 (P < 0.001) and rs = 0.717 (P < 0.001) for CD44 and EZR respectively. EZR also showed a significant increase from normal to pre-cancerous lesions (21% to 44%, P < 0.05). Similarly, no differences within SCC subgroups were detected for both proteins (Table 2). A positive correlation was found between CD44 and EZR, with rs= 0.473 (P < 0.001). Reputed as a cancer stem cell marker, strong labelling of CD44 discriminates SCC from BCC [35,36].

Table 2.

IHC staining of cell adhesion molecules CD44 and EZR in clinical samples

			Intensity of CD44/EZR staining
			-	+	++	+++
Normal	42	CD44	35 (83%)	2 (5%)	5 (12%)	0 (0%)	# ∧
		EZR	33 (79%)	8 (19%)	1 (2%)	0 (0%)	# ∧
Pre-cancerous	34	CD44	24 (70%)	3 (9%)	4 (12%)	3 (9%)	#
		EZR	19 (56%)	11 (32%)	4 (12%)	0 (0%)	#
Bowens disease	3	CD44	0 (0%)	1 (33%)	0 (0%)	2 (67%)
		EZR	1 (33%)	1 (33%)	1 (33%)	0 (0%)
Actinic keratosis	24	CD44	21 (88%)	0 (0%)	2 (8%)	1 (4%)
		EZR	15 (63%)	8 (33%)	1 (4%)	0 (0%)
Cutaneous horns	7	CD44	3 (42%)	2 (29%)	2 (29%)	0 (0%)
		EZR	3 (42%)	2 (29%)	2 (29%)	0 (0%)
SCC	51	CD44	20 (39%)	15 (29%)	5 (10%)	11 (22%)
		EZR	2 (4%)	16 (31%)	15 (30%)	18 (35%)
Well-differentiated	18	CD44	8 (45%)	4 (22%)	2 (11%)	4 (22%)
		EZR	1 (6%)	9 (50%)	6 (33%)	2 (11%)	¥
Moderately-differentiated	13	CD44	3 (23%)	5 (39%)	2 (15%)	3 (23%)
		EZR	0 (0%)	1 (8%)	4 (30%)	8 (62%)
Poorly-differentiated	20	CD44-	9 (45%)	6 (30%)	1 (5%)	4 (20%)
		EZR	1 (5%)	6 (30%)	5 (25%)	8 (40%)

#: Significantly different from SCC samples (P ≤ 0.05); ¥: Significantly different from the moderately-differentiated samples (P ≤ 0.05); ∧: Positively correlated with cancer initiation (P ≤ 0.05).

The membrane expressions of E-cadherin and β-catenin, the other two cell-cell junction proteins examined, were significantly reduced along in pathologic development of SCC (Figure 1C). A significant decrease of E-cadherin was detected between normal and SCC (98% to 92%, P < 0.001), pre-cancerous and SCC groups (100% to 92%, P < 0.001). Interestingly, although the changes in total number of positive staining were not dramatic, the intensity of E-cadherin staining was gradually shifted from predominantly very strong (+++, 88%) in normal tissues to only 20% (+++) in SCC group, with the majority staining significantly weaker, albeit not completely disappeared (Table 3). Akin to E-cadherin, the membrane β-catenin expression was significantly reduced from normal to SCC (93% to 69%, P < 0.001), and from normal to pre-cancerous lesions (93% to 82%, P = 0.001). The membrane abundance of both E-cadherin and β-catenin were inversely correlated with cancer initiation and progression (rs = -0.619, P < 0.001 and rs = -0.307, P < 0.001, respectively). However, no significant difference was found between SCC subgroups.

Table 3.

IHC staining of cell-cell junction molecules membranous E-cadherin and β-catenin in clinical samples

			Intensity of membrane E-cadherin/β-catenin staining
			-	+	++	+++
Normal	42	E-cadherin	1 (2%)	0 (0%)	4 (10%)	37 (88%)	# ∨
		β-catenin	3 (7%)	1 (2%)	29 (69%)	9 (22%)	# ∨
Pre-cancerous	34	E-cadherin	0 (0%)	0 (0%)	8 (24%)	26 (76%)	#
		β-catenin	6 (18%)	2 (6%)	25 (73%)	1 (3%)	*
Bowen’s disease	3	E-cadherin	0 (0%)	0 (0%)	0 (0%)	3 (100%)
		β-catenin	1 (33%)	0 (0%)	2 (67%)	0 (0%)
Actinic keratosis	24	E-cadherin	0 (0%)	0 (0%)	7 (29%)	17 (71%)
		β-catenin	5 (21%)	2 (8%)	16 (67%)	1 (4%)
Cutaneous horn	7	E-cadherin	0 (0%)	0 (0%)	1 (14%)	6 (86%)
		β-catenin	0 (0%)	0 (0%)	7 (100%)	0 (0%)
SCC	51	E-cadherin	4 (8%)	16 (31%)	21 (41%)	10 (20%)
		β-catenin	16 (31%)	8 (16%)	20 (39%)	7 (14%)
Well-differentiated	18	E-cadherin	0 (0%)	5 (28%)	9 (50%)	4 (22%)
		β-catenin	6 (33%)	3 (17%)	8 (44%)	1 (6%)
Moderately-differentiated	13	E-cadherin	3 (23%)	3 (23%)	4 (31%)	3 (23%)
		β-catenin	3 (23%)	4 (31%)	3 (23%)	3 (23%)
Poorly-differentiated	20	E-cadherin	1 (5%)	8 (40%)	5 (25%)	6 (30%)
		β-catenin	7 (35%)	1 (5%)	9 (45%)	3 (15%)

#: Significantly different from SCC samples (P ≤ 0.05); *: Significantly different from normal samples (P ≤ 0.05); ∨: Negatively correlated with cancer initiation (P ≤ 0.05).

β-catenin possesses the dual functions of intercellular communication on cell membranes and nuclear transactivator on Wnt pathway. The nuclear translocalization of β-catenin is implicated in various human cancer developments [37], including SCC [38], which dictates activation of downstream gene transcription to support cancer cell proliferation, progression and metastasis [39]. Indeed a correspondent increase of nuclear localization of β-catenin along with cancer formation was displayed (rs = 0.23, P < 0.01, data not shown), from normal to SCC (7% to 30%, P = 0.001, data not shown), and normal to pre-cancerous lesion groups (7% to 24%, P = 0.01, data not shown).

The expressions of the chaperone proteins Hsp75 and Hsp90-α were similarly increased from normal to SCC (Figure 1D), 7% to 73% (P < 0.001) and 29% to 100% (P < 0.001) respectively; from pre-cancerous and SCC groups, 15% to 73% (P < 0.001) and 65% to 100% (P < 0.05) respectively. Hsp75 and Hsp90-α were both positively correlated with cancer development, with rs = 0.593 (P < 0.001) and rs = 0.660 (P < 0.001) (Table 4). The Hsp90-α expression was also significantly higher in pre-cancerous than the normal group (29% to 65%, P = 0.001). It is notable that, whereas Hsp75 expression was not significantly different between the differentiation subgroups of SCC, the strength of Hsp90-α staining in SCC group was highly associated with differentiation phases (rs = 0.389, P = 0.005, Table 4). While Hsp90-α was predominantly (+) or (++) in the well and moderate-differentiated samples, it was strongly stained (+++) in half of the poorly-differentiated SCC samples.

Table 4.

IHC staining of the Chaperone proteins Hsp75 and Hsp90-α in clinical samples

			Intensity of Hsp75/Hsp90-α staining
			-	+	++	+++
Normal	42	Hsp75	39 (93%)	1 (2%)	2 (5%)	0 (0%)	# ∧
		Hsp90-α	30 (71%)	7 (17%)	3 (7%)	2 (5%)	# ∧
Pre-cancerous	34	Hsp75	29 (85%)	1 (3%)	1 (3%)	3 (9%)	#
		Hsp90-α	12 (35%)	6 (18%)	10 (29%)	6 (18%)	#
Bowen’s disease	3	Hsp75	3 (100%)	0 (0%)	0 (0%)	0 (0%)
		Hsp90-α	3 (100%)	0 (0%)	0 (0%)	0 (0%)
Actinic keratosis	24	Hsp75	22 (92%)	0 (0%)	0 (0%)	2 (8%)
		Hsp90-α	8 (33%)	6 (25%)	8 (33%)	2 (9%)
Cutaneous horn	7	Hsp75	4 (58%)	1 (14%)	1 (14%)	1 (14%)
		Hsp90-α	1 (14%)	0 (0%)	2 (29%)	4 (57%)
SCC	51	Hsp75	14 (27%)	8 (16%)	15 (30%)	14 (27%)
		Hsp90-α	0 (0%)	4 (8%)	33 (65%)	14 (27%)	◊
Well-differentiated	18	Hsp75	6 (33%)	3 (17%)	5 (28%)	4 (22%)
		Hsp90-α	0 (0%)	3 (17%)	12 (66%)	3 (17%)	&
Moderately-differentiated	13	Hsp75	3 (23%)	2 (15%)	5 (39%)	3 (23%)
		Hsp90-α	0 (0%)	1 (8%)	11 (84%)	1 (8%)	&
Poorly-differentiated	20	Hsp75	5 (25%)	3 (15%)	5 (25%)	7 (35%)
		Hsp90-α	0 (0%)	0 (0%)	10 (50%)	10 (50%)

#: Significantly different from SCC samples (P ≤ 0.05); &: Significantly different from the poorly-differentiated samples; ∧: Positively correlated with cancer initiation (P ≤ 0.05); ◊: Positively correlated with cancer cell differentiation levels (P ≤ 0.05).

The RNA-binding protein exosome component 10 (EXOSC10) has a role in mitotic cell division, mRNA biogenesis and stability, is implicated in clinical conditions such as autoimmune diseases [40] and was identified as a tumor suppressor gene in bladder carcinoma [41]. In our samples, its expression was significantly increased from normal to SCC (14% to 65%, P < 0.001) and from pre-cancerous to SCC (26% to 65%, P < 0.01) (Figure 1E; Table 5). The EXOSC10 staining was positively correlated with disease development (rs = 0.392, P < 0.001) (Table 5), but not related to cancer differentiation.

Table 5.

IHC staining of the RNA-binding protein EXOSC10 in clinical samples

	Intensity of EXOSC10 staining
		-	+	++	+++
Normal	42	36 (86%)	0 (0%)	5 (12%)	1 (2%)	# ∧
Pre-cancerous	34	25 (74%)	2 (6%)	4 (11%)	3 (9%)	#
Bowen’s disease	3	1 (33%)	0 (0%)	2 (67%)	0 (0%)
Actinic keratosis	24	21 (88%)	2 (8%)	0 (0%)	1 (4%)
Cutaneous horn	7	3 (42%)	0 (0%)	2 (29%)	2 (29%)
SCC	51	18 (35%)	18 (35%)	11 (22%)	4 (8%)
Well-differentiated	18	8 (44%)	7 (39%)	2 (11%)	1 (6%)
Moderately-differentiated	13	5 (39%)	2 (15%)	5 (39%)	1 (6%)
Poorly-differentiated	20	5 (25%)	9 (45%)	4 (20%)	2 (10%)

#: Significantly different from SCC samples (P ≤ 0.05); ∧: Positively correlated with cancer initiation (P ≤ 0.05).

Among mitochondrial proteins, the advance of cancer was accompanied by the progressively increased of SOD2 expression (Figure 1F). Statistical analysis revealed significant differences between normal and SCC (2% to 63%, P < 0.001), and pre-cancerous and SCC (0% to 63%, P < 0.001) groups. While the normal and pre-cancerous tissues were almost completely devoid of SOD2, its increased expression in SCC appeared abrupt and sharp (rs = 0.605, P < 0.001, Table 6), in a manner of correlation with the cancer differentiation status (rs = 0.423, P < 0.005).

Table 6.

IHC staining of the mitochondrial redox enzyme SOD2 in clinical samples

	Intensity of SOD2 staining
		-	+	++	+++
Normal	42	41 (98%)	0 (0%)	1 (2%)	0 (0%)	#
Pre-cancerous	34	34 (100%)	0 (0%)	0 (0%)	0 (0%)	# ∧
Bowen’s disease	3	3 (100%)	0 (0%)	0 (0%)	0 (0%)
Actinic keratosis	24	24 (100%)	0 (0%)	0 (0%)	0 (0%)
Cutaneous horn	7	7 (100%)	0 (0%)	0 (0%)	0 (0%)
SCC	51	19 (37%)	9 (18%)	14 (27%)	9 (18%)	◊
Well-differentiated	18	11 (61%)	4 (22%)	3 (17%)	4 (22%)
Moderately-differentiated	13	3 (25%)	2 (15%)	6 (50%)	2 (15%)	&
Poorly-differentiated	20	5 (25%)	3 (15%)	5 (25%)	7 (35%)

#: Significantly different from SCC samples (P ≤ 0.05); &: Significantly different from the poorly-differentiated samples; ∧: Positively correlated with cancer initiation (P ≤ 0.05); ◊: Positive correlated with cancer cell differentiation levels (P ≤ 0.05).

ICC validation of the biomarkers as therapeutic targets of resveratrol

According to the IHC staining of the resection specimens, the 10 candidate proteins (CK10, CK17, CD44, EZR, E-cadherin, β-catenin, Hsp75, Hsp90-α, EXOSC10 and SOD2) representing five functional aspects were marked out, to a substantial point, as the potential molecular footprints of squamous cell cancer initiation and progression. An immunocytochemistry method (ICC) was employed to profile their changes in the squamous cell carcinoma cell lines SCL12 and COLO16 upon treatment of resveratrol for 48 hrs. The proliferation rates of both cell lines were inhibited by 100 µM resveratrol, but COLO16 were more resveratrol-sensitive (data not shown). Whether the differential responses of the cells to resveratrol were reflected in the alteration of these selected proteins was tested.

The expression of CK10, which was found lost in squamous cell cancer tissue and cell lines, was restored by resveratrol treatment in both cell lines, but to a much greater extent in COLO16 cells. In contrast, CK17, the cytokeratin negatively correlated to CK10, was inhibited in COLO16 cells, but not in SCL12 cells (Figure 2A).

Figure 2.

Representative ICC images of SCL12 and COLO16 cells treated with resveratrol. After treatment with 100 µM resveratrol for 48 hours, cancer cells were stained with different antibodies. The representative results showed protein expressions of the candidate markers in groups of cytokertatins (CK10, CK17), cellular adhesion molecules (CD44, EZR), cell-cell junction proteins (membrane β-catenin, E-cadherin), chaperone proteins (Hsp90-α, Hsp75), RNA-binding protein (EXOSC10) and mitochondrial redox enzyme (SOD2) in normal, pre-cancerous or SCC patient samples.

Staining of the cell adhesion-related proteins CD44 and EZR, which arepositively correlated with cancer deterioration, was reduced in COLO16 cells when treated with resveratrol, but not in SCL12 cells (Figure 2B). Resveratrol increased the membrane expression of β-catenin in both cell lines, but E-cadherin was only increased in SCL12 cells (Figure 2B). In contrast, resveratrol treatment resulted in a decreased expression of Hsp90-α only in COLO16 cells, while no significant changes of Hsp75 were observed in both cells (Figure 2C). A remarkable reduction of EXOSC10 abundance was detected in COLO16 cells upon resveratrol treatment, but not the SCL12 cells (Figure 2D). The mitochondrial protein SOD2, which was strongly correlated with the pathological stages of cancer, was interestingly expressed at higher level in COLO16 cells after resveratrol treatment, but lower in SCL12 cells (Figure 2E). This may suggest a context-dependent effect of resveratrol in SCC.

Discussion

An important feature of SCC is its development in multiple steps, with common precursors including actinic keratoses (AKs, also known as solar keratoses) and squamous cell carcinoma in situ (Bowen’s disease), which in most cases results directly from excessive UV-damage and can present indolent or non-invasive. Given certain variations by age, gender, and geographic localization, almost half of the global populations suffers from AKs and its risk of progression into invasive SCC fluctuates from 0.1% to 20% [42]. Comparing the focal skin damage in AKs, Bowen’s disease affects the whole thickness of epidermis and is more frequently found in the elderly over 60. Despite the relatively low rate of invasive evolvement (3-5%) of Bowen’s disease to invasive SCC, a fifth of the progressed cases become metastatic [43]. Teasing out the decisive factors driving the characteristic development of SCC from benign to invasive is essential.

To varying degrees, all of the ten protein candidates have shown disease progression-specific significance. Cytokeratins are a group of epithelia-specific intermediate-sized filament proteins, a constitutive component of the cytoskeleton in vertebrate cells. Its implication in various cancers has been characterized long ago [44,45]. Together with other cytokeratins, higher CK17 is suggestive of moderate oral epithelial dysplastic changes [46]. Expression of CK17 has been proposed as a sensitive and specific biomarker for the malignant transformation of squamous carcinoma of the larynx [47]. In the same study, CK10 was identified as a specific indicator of cancer progression, which is in contrast to the present ICC result where an inverse correlation was found in our samples. The disagreement is probably due to the different origins of the cancer tissues. Furthermore, ICC results indicated a resveratrol-induced reversal of the pattern of CK10/CK17 expression in COLO16 cell line, but not in the resveratrol-resistant SCL12 cells, which may contribute to the sensitivity of COLO16 cells to resveratrol. CK is proven to possess a value of differential diagnosis between keratoacanthoma (KA) and SCC in an analysis of biopsy samples where CK10 staining was rather scattered in small groups of cells of SCC, but in 98% of KAs [48], which reinforces our conclusion that loss of CK10 is a specific sign of invasive SCC.

The cell membrane molecule CD44 is referred to as one of the cancer stem cell (CSC) markers and found overexpressed in different tumors, dictating higher levels of cellular self-renewal and leading to cancer aggressiveness and reoccurrence. Despite the heterogeneity of CD44 expression in diverse cell populations, an association between CD44 expression and worse overall survival was reported in tongue SCC patients [49]. As a member of erzrin/radixin/moesin (ERM) protein family that provides cortical actin cytoskeleton support of membrane architecture and molecular activities, ezrin forms a close structural and functional link with CD44 to control normal cell-cell communication and mediate tumor-endothelium interactions. Their co-expression has been found in normal membrane ruffles and metastatic cells [50,51], which is in good agreement with our data showing a parallel increase of expressions of CD44 and EZR in cancerous tissues. The CD44/EZR may form a complex that could be a target of resveratrol treatment, as shown in present ICC results from COLO16 cells.

Another pair of membrane proteins with a structural association, E-cadherin and β-catenin, was also down-regulated in the bSCC examined, in line with the literature. But instead of being a progressive reduction [52], significant fainting of E-cadherin staining was observed in SCC, but no differences between normal and pre-cancerous samples, suggesting that E-cadherin might be indicative of gaining of a certain trait of invasiveness. Correspondingly the nuclear abundance of β-catenin was increased along the histopathologic development of cancer, implying its translocation to facilitate its transcriptional activation in cancer. Similar results were reported in the study of oral squamous cell cancer [53]. Upon treatment with resveratrol, the membrane β-catenin was recovered only in COLO16 cells, suggesting that its intracellular localisation may play a part in the therapeutic effects of resveratrol.

Although positive or inverse correlations between protein expression and cancer development were found in all proteins, the chaperone proteins, Hsp75 and Hsp90-α, and SOD2 were the only three that had a correlation with differentiation stages of cancer. The chaperone protein expressions increased in parallel at each step of SCC progression, whereas SOD2 had no implication in incipient skin damage, as its expression was similar in normal and pre-cancerous samples. Aberrant expression of Hsp90-α is found in melanoma and reagents inhibiting its C- or N-terminal have been tested to curb the aggressive development of melanoma [54]. Using a proteomic approach, Hsp75 expression was elevated and oxidatively modified in HPV-transformed human keratinocyte HK-168 cells upon UVB irradiation [55]. The dysregulation and oxidation of the Hsp family are linked to a higher level of protein misfolding and misfunction, and aid tumor growth by perturbance in autonomous cell growth and escape from programmed cell death, which are characteristic of cancer. Increased Hsp has been shown to participate in the microenvironment conditioning to favor cancer evolution, which makes it a promising target in tumor therapy [56]. In our study, a reduced Hsp90-α expression by resveratrol was observed in COLO16 cells, which may contribute to its pro-apoptotic effect in the cells.

The prerequisite of Hsp70 for the survival of cancer cells against apoptotic and necrotic stimuli has been proven in solid carcinomas of breast, colon, prostate, liver, and glioblastoma [1,57]. As a member of the Hsp70 family, Hsp75 is critical for the importation of cytosolic SOD2 to mitochondria where SOD executes its antioxidant functions [58,59]. Hsp75 is inducible by oxygen-glucose deprivation, oxidative stress, and UVA irradiation, the hallmarks of cancers [60]. In both cell lines tested in our study, resveratrol did not bring down the enhanced expressions of Hsp75 and SOD2. It is not clear whether there is a cooperative mechanism between these two proteins.

Failure to repair the DNA double-strand breaks (DSBs) arising from excessive UV exposure is the primary detrimental factor for skin damage. As an important member of mRNA surveillance machinery, the exosome EXOSC10 plays a necessary role in DNA repair through direct recruitment to the sites of DSBs induced by radiation and interaction with RAD51 [61]. The positive correlation between EXOSC10 and SCC progression found in this study might suggest a feedback mechanism to maintain genomic integrity. It will be interesting to examine whether it functions to promote or suppress cancer development.

Cancer cells can adapt strategies to create a fostering niche to achieve maximum survival and proliferation, including metabolic adjustment, cellular dedifferentiation, and escape from immune surveillance. Therapeutic reagents targeting the whole reprogramming of the microenvironment, especially at the early stage, is probably the best way to assure a brighter prognosis.

The bio-signature of proteins not only varies at each step of SCC development, but also is dependent on the etiologic factors and the exact location of cancer occurrence. For example, UV light-induced SCC may display different protein profiles compared with SCC caused by chemicals or viral infection; SCC lesions residing in the skin may differ from oral cavity, lung, or bladder [1].

It is admitted that the present study has a limitation of the small sample size of different AKs and Bowen’s disease, which may explain why there were none of the significant changes of the protein candidates found within these groups. In addition, although the pro-neoplastic inflammatory infiltrates in intraepithelial lesions have been proposed as an important promoter of SCC [62-64], we did not discern a significant pattern of inflammation signature in our study (data not shown). This may be because the material we used did not cover clinical samples or a wider range of SCC cells. These considerations will surely point us toward future investigations.

Disclosure of conflict of interest

None.