This article is mentioned in:

Introduction

Oral squamous cell carcinoma (OSCC) represents the sixth most common type of cancer worldwide (1). Several biomarkers for OSCC have been demonstrated previously (2–7) and may be useful for the diagnosis and prognosis of oral cancer. As carcinogenesis is a multistep process, a number of genes involved in the diagnosis of oral cancer have not been identified.

Integrins are a family of transmembrane-type receptor proteins on the cell surface, composed of heterodimeric complexes of 1α chain and 1β chain. The 18α and 8β subunits comprise ~24 different integrin receptors, each of which is capable of binding to a specific type of cell surface and extracellular matrix (ECM) protein ligand (8). They function in specific signal transduction and adhesive interactions between cells, and between cells and the ECM. Integrin signaling may modulate several different functions involved in cell motility, invasion, proliferation and migration (9,10). Furthermore, integrin signaling may also affect vascular neogenesis (11). Therefore, integrins serve an important role in tumor growth and metastasis (12–14).

Integrin subunit α3 (ITGA3), which combines with integrin subunit β1 (ITGB1) to form integrin α3β1, is the receptor for ECM molecules including fibronectin, laminin and collagen (8,12). Integrin α3β1 has been demonstrated to function in cell proliferation, migration and motility, and in the maintenance of basement membrane integrity (15–18). Several previous studies identified that integrin α3β1 was overexpressed and associated with tumor invasion and metastasis in the majority of types of cancer, including lung (19) and breast (20) cancer cell lines. Integrin subunit α5 (ITGA5) often combines with ITGB1 to form integrin α5β1, and serves as a receptor for fibronectin and fibrinogen to participate in cell differentiation, cell development and wound healing (8,12). It was identified that the emergence of integrin α5β1 expression is associated with tumor progression in lung cancer (21). ITGA5 may also promote tumor metastasis in oral cancer cell lines (22). Integrin subunit β6 (ITGB6) is the β subunit of integrin αvβ6, which is a receptor for fibronectin and cytotactin, and regulates cell invasion, inhibits apoptosis, modulates matrix metalloproteases and activates transforming growth factor β1 (8,23). Overexpression of integrin αvβ6 promotes epithelial-to-mesenchymal transition and is associated with cell invasion in oral cancer cell lines (24,25). However, the majority of these previous studies were focused on OSCC cell line models or non-quantitative immunohistochemical analyses of OSCC tissues. These studies did not use receiver-operating characteristic (ROC) curve analyses, which provide sensitivity and specificity data for the evaluation of OSCC biomarker performances.

In the present study, the reverse transcription-quantitative polymerase chain reaction (RT-qPCR) analysis was used to determine the mRNA expression levels of ITGA3, ITGA5, ITGB1 and ITGB6 genes in OSCC tissues from different tumor locations for ROC curve analyses, in order to identify suitable biomarker genes for the diagnosis of early-stage OSCC. The mRNA expression of ITGA3, ITGA5, ITGB1, and ITGB6 genes were identified as potential OSCC biomarkers. Furthermore, cumulative ROC analysis of integrin OSCC biomarkers had an improved diagnostic performance compared with individual ROC analysis.

Materials and methods

Tissue samples

The present study was approved by the Institutional Review Board at Kaohsiung Medical University (Kaohsiung, Taiwan) (approval no. KMUH-IRB-930104), and all patients provided written informed consent. A total of 55 oral tumors and 55 matched normal oral control tissues (at least 2.5 cm between tumor and control tissues) were collected (December 2004 to December 2009) from the Department of Oral and Maxillofacial Surgery, Kaohsiung Medical University Hospital. The age range of this patient cohort is 30-90 years and the median age is 50 years. All samples were blindly examined by at least three pathologists of the Department of pathology, Kaohsiung Medical University Hospital. All control tissues underwent pathological diagnosis for confirmation as non-tumor. All oral tumors underwent pathological diagnosis for OSCC and tumor stage classification used the Tumor-Node-Metastasis (TNM) system (26). The characteristics of the patients with OSCC are summarized in Table I, this basic patient information has been described previously (6).

Table I. Basic characteristics of patients and tissue samples.

Table I.

Basic characteristics of patients and tissue samples.

	(OSCC/matched control oral tissue)
Characteristics	n (total=55)	(%)
Age, years
≤40	8	14.55
41-50	21	38.18
51-60	21	38.18
≥61	5	9.09
Sex
Male	50	90.91
Female	5	9.09
TNM stage
I	10	18.18
II	9	16.36
III	16	29.09
IV	20	36.37
Lesion location
Buccal mucosa/retromolar area	23	41.82
Tongue/mouth floor	17	30.91
Edentulous ridge	8	14.55
Others	7	12.72
Carcinogenic factors
Alcohol
(+)	46	83.64
(−)	9	16.36
Betel quid
(+)	50	90.91
(−)	5	9.09
Cigarette smoking
(+)	48	87.27
(−)	7	12.73

[i] This basic patient information has been described previously (6). TNM, Tumor-Node-Metastasis; Others, lower lip/vestibule/soft palate; OSCC, oral squamous cell carcinoma; +, yes; -, no.

RT-qPCR

Tissues were sliced and placed into 1.5 ml microcentrifuge tubes containing TRIzol® reagent (Thermo Fisher Scientific, Inc., Waltham, MA, USA) for homogenization using a disposable tissue grinder pestle. Subsequently, the homogenized mixtures were used for total RNA extraction according to the manufacturer's protocol. RT was performed using an OmniScript RT kit (Qiagen GmbH, Hilden, Germany). qPCR was performed using an iQ SYBR Green Supermix (Bio-Rad Laboratories, Inc., Hercules, CA, USA) in an iCycler MyiQ real-time machine (Bio-Rad Laboratories, Inc.). The relative mRNA expressions of ITGA3, ITGA5, ITGB1 and ITGB6 genes in OSCCs and controls were examined. The forward and reverse primer sequences, and the lengths of PCR products, for these genes are provided in Table II. The thermocycling conditions are as follows: 94°C for 1 min; 4 cycles of 94°C for 15 sec, 64°C for 15 sec and 70°C for 15 sec; 4 cycles of 94°C for 15 sec, 61°C for 15 sec and 70°C for 15 sec; 4 cycles of 94°C for 15 sec, 58°C for 15 sec and 70°C for 15 sec; 60 cycles of 94°C for 15 sec, 55°C for 15 sec and 70°C for 15 sec; and finally 94°C for 1 min and 60°C for 5 min. The PCR assay was performed in duplicate. The relative mRNA expression levels (OSCC/oral tissue from the tumor-free margin ratio) for ITGA3, ITGA5, ITGB1 and ITGB6 genes were evaluated using the 2−ΔΔCq method (27). The threshold cycle (Cq) value of each integrin gene was subtracted from the Cq value for the GAPDH reference housekeeping gene. Melting curve analyses by qPCR analysis and gel electrophoresis by 1.5% agarose containing 0.5 µg/ml ethidium bromide for visualization under a UV box were used to validate the PCR products as described previously (28).

Table II. Forward and reverse primer sequences and the lengths of PCR products for ITGA3, ITGA5, ITGB1, ITGB6, and GAPDH genes.

Table II.

Forward and reverse primer sequences and the lengths of PCR products for ITGA3, ITGA5, ITGB1, ITGB6, and GAPDH genes.

Gene names	Primer sequences	Length of PCR products, bp
ITGA3	Forward: 5′-TGCTGTGGAAGTGCGGCT-3′	206
	Reverse: 5′-GCGTGGTACTTGGGCATGAT-3′
ITGA5	Forward: 5′-TCATCTACATCCTCTACAAGCTTGG-3′	204
	Reverse: 5′-GCCGTCAGCACCTTCAAGA-3′
ITGB1	Forward: 5′-CGTATTCAGTGAATGGGAACAAC-3′	231
	Reverse: 5′-GATTTTCACCCGTGTCCCAT-3′
ITGB6	Forward: 5′-ACATGAAAGTGGGAGACACAGC-3′	215
	Reverse: 5′-ACACACCCCACACTGGAAAGA-3′
GAPDH	Forward: 5′-GCATCCTGGGCTACACTGA-3′	162
	Reverse: 5′-CCACCACCCTGTTGCTGTA-3′

[i] ITGA3, integrin subunit α3; ITGA5, integrin subunit α5; ITGB1, integrin subunit β1; ITGB6, integrin subunit β6.

Statistical analysis

Data were analyzed by the Mann-Whitney test and Kruskal-Wallis test using SPSS version 17.0 software (SPSS, Inc., Chicago, IL, USA). ROC curve analyses were commonly used to evaluate the performance of cancer diagnosis (29–34). Accordingly, ROC curves were used to examine the performance of ITGA3, ITGA5, ITGB1 and ITGB6 genes as predictive biomarkers for detecting OSCC. Cut-off values for individual ROC curves were calculated using the 60-ΔCq value, where 60 represents the total PCR cycle number, and no signal was assigned to the Cq value for 60, as described previously (5,6). For individual ROC analysis, the area under the ROC curve (AUC) of each biomarker was used to evaluate its performance as an OSCC biomarker.

To consider the combined effect of each biomarker, the cumulative ROC analysis of biomarkers was performed for OSCC prediction. At first, the critical (cut-off) point of the ROC curve for each biomarker was identified using JMP version 10 statistic software (SAS Institute Inc., Cary, NC, USA). Subsequently, biomarkers that were more relevant to the OSCC (those with values greater than the cut-off point in AUC of individual ROC results) were assigned a score of 1 (for example, values greater than the cut-off point from the AUC analysis of the individual ROC results), and the remaining biomarkers were assigned a score of 0. The combined scores for different combinations of biomarkers were calculated using the formula function in JMP 10. Finally, the combined scores were used for cumulative ROC analysis of biomarkers for comparison.

Results

Comparison of clinicopathological features and relative mRNA expressions of ITGA3, ITGA5, ITGB1 and ITGB6 genes in patients with OSCC

To evaluate the performance of the OSCC biomarkers, mRNA expression levels of ITGA3, ITGA5, ITGB1 and ITGB6 genes of OSCC samples were evaluated by comparing tumor samples with control samples (oral tissues from the tumor-free margin). In different locations (Table III), the differences in ITGA3 and ITGB6 gene expression levels were significant (P=0.05 and 0.005, respectively by Kruskal-Wallis test); however, the differences in the ITGA5 and ITGB1 gene expression levels were non-significant. In contrast, the differences in mRNA expression levels of ITGA3, ITGA5, ITGB1 and ITGB6 genes were non-significant at different stages (I–IV) (P=0.19, 0.58, 0.92, and 0.53, respectively by Kruskal-Wallis test) and for the T parameter (1–4) of the TNM staging system (a measure of tumor diameter/dimension) among patients with OSCC (P=0.36, 0.97, 0.75, and 0.83, respectively by Kruskal-Wallis test).

Table III. Comparison of clinicopathological features and relative mRNA expression levels of ITGA3, ITGA5, ITGB1 and ITGB6 genes in oral squamous cell carcinoma tumor samples with control samples (oral tissue from the tumor-free margin).

Table III.

Comparison of clinicopathological features and relative mRNA expression levels of ITGA3, ITGA5, ITGB1 and ITGB6 genes in oral squamous cell carcinoma tumor samples with control samples (oral tissue from the tumor-free margin).

		ITGA3		ITGA5		ITGB1		ITGB6
Clinical data	n (total=55)	Mean ± SD	P-valuea	Mean ± SD	P-valuea	Mean ± SD	P-valuea	Mean ± SD	P-valuea
Locationb			0.01		0.33		0.16		0.05
1	23	1.95±5.37		1.29±4.87		2.58±11.30		0.71±6.19
2	17	4.32±4.74		4.04±4.99		2.41±3.23		0.91±11.08
3	8	6.25±5.20		1.82±8.40		2.94±3.03		5.07±5.30
4	7	−6.87±14.91		−1.74±12.27		2.15±3.58		2.66±5.95
TNM			0.19		0.58		0.92		0.53
T1	14	1.52±12.93		2.88±5.64		1.20±4.29		−2.12±13.83
T2	21	1.04±4.84		0.65±7.58		1.29±3.66		1.67±4.95
T3 and T4	20	3.86±5.48		2.34±6.82		4.74±11.37		2.66±3.69
Stage			0.36		0.97		0.75		0.83
I	10	8.26±15.43		2.53±5.56		0.87±5.03		0.16±9.95
II	9	0.30±5.13		2.10±4.09		0.83±4.47		0.83±5.86
III and IV	36	3.03±4.99		1.57±7.07		3.41±8.70		1.37±8.02

{ label (or @symbol) needed for fn[@id='tfn3-ol-0-0-9168'] } Data are presented as mean ΔΔCq values (ΔCq of OSCC-ΔCq of its matched oral tissue from the tumor-free margin). TNM, Tumor-Node-Metastasis; T in TMN, tumor size and invasiveness in TMN classification.

a Kruskal-Wallis test.

b Location 1, buccal mucosa/retromolar area; 2, tongue/mouth floor; 3, edentulous ridge; 4, others (lower lip/vestibule/soft palate). SD, standard deviation; ITGA3, integrin subunit α3; ITGA5, integrin subunit α5; ITGB1, integrin subunit β1; ITGB6, integrin subunit β6.

Comparison of patient habits and relative mRNA expression of ITGA3, ITGA5, ITGB1, and ITGB6 genes in patients with OSCC

As indicated in Table IV, the differences in mRNA expression levels of ITGA3, ITGA5, ITGB1 and ITGB6 genes were not associated with cancer-causing habits, including alcohol consumption (P=0.83, 0.11, 0.68, and 0.18, respectively), betel nut chewing (P=0.65, 0.55, 0.81, and 0.75, respectively) and cigarette smoking (P=0.84, 0.45, 0.86, and 0.40, respectively by Mann-Whitney test).

Table IV. Comparison of patient habits and relative mRNA expression levels of ITGA3, ITGA5, ITGB1 and ITGB6 genes in oral squamous cell carcinoma samples with control samples (oral tissue from the tumor-free margin).

Table IV.

Comparison of patient habits and relative mRNA expression levels of ITGA3, ITGA5, ITGB1 and ITGB6 genes in oral squamous cell carcinoma samples with control samples (oral tissue from the tumor-free margin).

		ITGA3		ITGA5		ITGB1		ITGB6
Habits	n (total=55)	Mean ± SD	P-valuea	Mean ± SD	P-valuea	Mean ± SD	P-valuea	Mean ± SD	P-valuea
Alcohol			0.83		0.11		0.68		0.18
No	9	3.02±3.60		4.66±3.38		1.91±2.26		4.53±5.73
Yes	46	2.02±8.42		1.27±7.18		2.65±8.27		0.39±8.22
Betel quid chewing			0.65		0.33		0.81		0.75
No	5	3.59±5.84		4.25±5.21		0.73±3.33		3.74±7.03
Yes	50	2.05±8.02		1.59±6.93		2.70±7.90		0.80±8.08
Cigarette smoking			0.84		0.45		0.86		0.40
No	5	3.51±4.90		3.53±6.61		1.13±2.31		4.56±6.98
Yes	50	2.05±8.07		1.66±6.86		2.66±7.94		0.71±8.05

{ label (or @symbol) needed for fn[@id='tfn6-ol-0-0-9168'] } Data are presented as mean ΔΔCq values (ΔCq of OSCC-ΔCq of its matched oral tissue from the tumor-free margin).

a Mann-Whitney U test. SD, standard deviation; ITGA3, integrin subunit α3; ITGA5, integrin subunit α5; ITGB1, integrin subunit β1; ITGB6, integrin subunit β6.

AUC performances of individual OSCC biomarkers of ITGA3, ITGA5, ITGB1, and ITGB6 genes

To test the predictive performance of each of the test biomarkers in OSCC diagnosis, individual ROC curves for ITGA3, ITGA5, ITGB1 and ITGB6 mRNA expression were constructed by comparing the mRNA expression level between OSCC tumor tissues and their controls (oral tissue from the tumor-free margin). The individual AUCs of the relative mRNA expression for ITGA3, ITGA5, ITGB1 and ITGB6 genes were 0.724 [95% confidence interval (CI), 0.630-0.819], 0.698 (95% CI, 0.600-0.796), 0.640 (95% CI, 0.536-0.743) and 0.657 (95% CI, 0.555-0.759), respectively (Fig. 1A). At a specificity of 50.91% for all locations (Table V, part A), sensitivities of relative mRNA expression for ITGA3, ITGA5, ITGB1 and ITGB6 were 90.91, 83.64, 70.91 and 74.55%, respectively. At a specificity of 61.82%, sensitivities of relative mRNA expression for ITGA3 and ITGA5 genes were decreased slightly to 85.45 and 78.18%, respectively. However, ITGB1 and ITGB6 genes were decreased to 50.91 and 61.82%, respectively. At a specificity of 67.27%, sensitivities of relative mRNA expression for ITGA3, ITGA5, ITGB1 and ITGB6 genes were markedly decreased, particularly for the ITGB1 gene (45.45%). Therefore, analysis of different sensitivities suggested that the ITGA3 and ITGA5 genes exhibited an improved specificity performance compared with ITGB1 and ITGB6 genes for all locations.

Figure 1. AUC values of ITGA3, ITGA5, ITGB1 and ITGB6 genes as individual oral squamous cell carcinoma biomarkers. (A) All tumor locations (locations 1-4). (B) Only tumor locations 2/3. Location 1, buccal mucosa/retromolar area; location 2, tongue/mouth floor; location 3, edentulous ridge; location 4, others (lower lip/vestibule/soft palate); ITGA3, integrin subunit α3; ITGA5, integrin subunit α5; ITGB1, integrin subunit β1; ITGB6, integrin subunit β6; AUC, area under the curve.

Table V. Different cut-offs and their relative sensitivity and specificity for ITGA3, ITGA5, ITGB1 and ITGB6 genes in patients with oral squamous cell carcinoma.

Table V.

Different cut-offs and their relative sensitivity and specificity for ITGA3, ITGA5, ITGB1 and ITGB6 genes in patients with oral squamous cell carcinoma.

A, All locations
	ITGA3		ITGA5		ITGB1		ITGB6
Sensitivity, %	Specificity, %	Cut-off	Specificity, %	Cut-off	Specificity, %	Cut-off	Specificity, %	Cut-off
50.91	90.91	63.20	83.64	56.26	70.91	61.93	74.55	57.90
61.82	85.45	62.87	78.18	55.00	50.91	61.17	61.82	57.21
67.27	67.27	61.94	63.64	54.35	45.45	60.96	58.18	57.06
B, Locations 2/3
	ITGA3		ITGA5		ITGB1		ITGB6
Sensitivity, %	Specificity, %	Cut-off	Specificity, %	Cut-off	Specificity, %	Cut-off	Specificity, %	Cut-off
64.00	92.00	62.97	84.00	55.00	64.00	61.07	84.00	57.90
68.00	88.00	62.67	84.00	55.00	64.00	60.99	80.00	57.65
76.00	76.00	61.73	76.00	53.94	60.00	60.78	60.00	56.61

[i] All locations including the locations 1, 2, 3 and 4. Location 1, buccal mucosa/retromolar area; location 2, tongue/mouth floor; location 3, edentulous ridge; location 4, others (lower lip/vestibule/soft palate); ITGA3, integrin subunit α3; ITGA5, integrin subunit α5; ITGB1, integrin subunit β1; ITGB6, integrin subunit β6.

As the differences in ITGA3 and ITGB6 gene expression by location were significant (Table III), whether the different locations of OSCC may affect the mRNA expression levels of these four integrin genes was additionally investigated. In the example of locations 2 and 3 combined (locations 2/3), i.e., the tongue/mouth floor and edentulous ridge (Fig. 1B), it was demonstrated that the AUC values of relative mRNA expression for ITGA3, ITGA5, ITGB1 and ITGB6 genes in the tongue/mouth floor and edentulous ridge were 0.840 (95% CI, 0.728-0.952), 0.765 (95% CI, 0.632-0.898), 0.725 (95% CI, 0.584-0.866) and 0.762 (95% CI, 0.630-0.896), respectively. Notably, the AUC values of locations 2/3 combined (Fig. 1B) for these four integrin genes were increased compared with those of all locations together (Fig. 1A). Analysis of different sensitivities suggested that ITGA3 and ITGA5 genes exhibited an improved specificity performance compared with that of ITGB1 and ITGB6 genes for locations 2/3 (Table V, part B).

AUC performances of cumulative ROC analyses for different combinations of ITGA3, ITGA5, ITGB1, and ITGB6 biomarkers

To evaluate the diagnostic power of different combinations of these biomarkers, their cumulative ROC curves between OSCC and controls (oral tissue from the tumor-free margin) were calculated. The AUC values for different combinations (two, three and four) of the biomarkers are summarized in Table VI. The combination of ITGA3, ITGA5 and ITGB1 genes demonstrated the highest AUC values of 0.809 (CI, 0.728-0.890) and 0.871 (CI, 0.770-0.972), for the two types of locations (all locations and locations 2/3, respectively). The combination of ITGA3, ITGA5, ITGB1 and ITGB6 genes provided similar AUC values for the two locations.

Table VI. Performances of cumulative ROC analyses for different combinations of ITGA3, ITGA5, ITGB1 and ITGB6 biomarkers in patients with OSCC.

Table VI.

Performances of cumulative ROC analyses for different combinations of ITGA3, ITGA5, ITGB1 and ITGB6 biomarkers in patients with OSCC.

	AUC of combined biomarkers for OSCCa
	AUC of combined biomarkers for OSCCa
Locations	ITGA3/ITGA5	ITGA3/ITGB1	ITGA3/ITGB6	ITGA5/ITGB1	ITGA5/ITGB6	ITGB1/ITGB6	ITGA3/ITGA5/ITGB1	ITGA3/ITGA5/ITGB6	ITGA5/ITGB1/ITGB6	ITGA3/ITGA5/ITGB1/ITGB6
Allb	0.777	0.775	0.758	0.762	0.720	0.686	0.809c	0.780	0.750	0.797
2/3	0.847	0.829	0.850	0.823	0.792	0.788	0.871c	0.855	0.825	0.870

a Biomarkers that were more relevant to the OSCC (those with values greater than the cut-off point in AUC of individual ROC results, as demonstrated in Fig. 1), were assigned a score of 1, and the others were assigned a score of 0. The combined scores for different combinations of biomarkers were used for cumulative ROC analysis.

b All locations including the locations 1, 2, 3, and 4.

c The best AUC performance among different combinations of biomarkers for OSCC. Location 1, buccal mucosa/retromolar area; location 2, tongue/mouth floor; location 3, edentulous ridge; location 4, others (lower lip/vestibule/soft palate); AUC, area under the curve; ROC, receiver operating characteristics; ITGA3, integrin subunit α3; ITGA5, integrin subunit α5; ITGB1, integrin subunit β1; ITGB6, integrin subunit β6; OSCC, oral squamous cell carcinoma.

Discussion

The overexpression of ITGA3, ITGA5, ITGB1 and ITGB6 genes has been demonstrated in several types of cancer (19–25); however, the suitability for ITGA3, ITGA5, ITGB1 and ITGB6 genes as OSCC biomarkers has rarely been considered.

The results of the present study identified that ITGA3 and ITGB1 mRNA were overexpressed in patients with OSCC. Similarly, it was demonstrated previously that ITGA3 and ITGB1 proteins were overexpressed in prostate tumor tissues, and knockdown of ITGA3 and ITGB1 genes by small interfering RNAs inhibited cell migration and invasion in prostate cancer cells (35). In the present study, ITGA5 mRNA was overexpressed in patients with OSCC. Similarly, ITGA5 and ITGB1 genes have been suggested to be potential biomarkers for non-small cell lung cancer (36). In addition, the ITGB6 gene may be a prognostic biomarker for invasive breast cancer (37). However, to the best of our knowledge, these integrin biomarkers have rarely been investigated in OSCC.

In the present study, individual ROC analyses for all locations identified ITGA3 and ITGA5 genes as good biomarkers for OSCC (AUC=0.724 and 0.698, respectively), but ITGB1 and ITGB6 genes performed poorly for OSCC prediction (AUC <0.66). Compared with all locations, improved AUC values for ITGA3, ITGA5, ITGB1 and ITGB6 genes were observed in combined locations 2 and 3 (the tongue/mouth part and edentulous ridge).

Previous studies have indicated that a combination of several variants generally provides an improved diagnostic power for ROC analysis compared with ROC analysis of individual markers (38–41). A combination of several biomarkers may therefore improve the diagnostic results for tumors, as has been demonstrated in thyroid cancer (42). Similarly, with the exception of the combination of ITGB1 and ITGB6 (AUC=0.686), the present study identified that the majority of the biomarker combinations, including a combination of two (ITGA3/ITGA5, ITGA3/ITGB1, ITGA3/ITGB6, ITGA5/ITGB1 and ITGA5/ITGB6), three (ITGA3/ITGA5/ITGB1, ITGA3/ITGA5/ITGB6 and ITGA5/IGB1/ITGB6) and four (ITGA3/ITGA5/ITGB1/ITGB6) genes, exhibited improved AUC values for all locations (ranging between 0.750 and 0.809) compared with the individual AUC values of individual integrin genes (ranging between 0.640 and 0.724). With the exception of the combinations of ITGA3/ITGB1, ITGA5/ITGB1, ITGA5/ITGB6, ITGB1/ITGB6 and ITGA5/ITGB1/ITGB6 genes (AUC values ranging between 0.788 and 0.829), it was identified that other combined biomarkers, including two, three and four genes, demonstrated improved AUC values for locations 2 and 3 (ranging between 0.847 and 0.871) compared with that of the individual AUC values of these integrin genes (ranging between 0.725 and 0.840). Among them, ITGA3, ITGA5 and ITGB1 genes exhibited the highest AUC values for all locations and locations 2 and 3 combined. Therefore, the cumulative ROC analysis, as a method of sensitive and specific evaluation, suggested a combination of multiple biomarkers for the diagnosis of OSCC.

mRNA has been repeatedly demonstrated to be a reliable material for the diagnosis of oral cancer (5,6,43,44). For example, tissue and salivary mRNA biomarkers are suggested to be associated with clinicopathological parameters for the diagnosis of OSCC (44). In clinical practice, the data from the present study of combined integrin biomarkers may be applied for non-invasive diagnosis of OSCC using saliva material for OSCC in the future.

An advantage of the proposed methods of the present study is a potential improvement of the AUC performance by using accumulative ROC analysis for the mRNA expression of a combination of integrin biomarkers. A disadvantage of this method is that it is based on tissue samples for mRNA evaluation, which are susceptible to degradation by RNase contamination (45). Additionally, expression at mRNA transcriptome levels will not always be consistent with those at protein levels in cases of posttranslational modification (46). Additionally, the present study did not provide information concerning depth of invasion, which is important for clinicians. Additional protein evaluation for ITGA3, ITGA5, ITGB1 and ITGB6 using immunohistochemistry is warranted. In addition, 5/55 the patients included in the present study were female, which may provide a gender bias for these biomarker predictions for oral cancer. Therefore, it is suggested that a larger cohort of female and male patients should be studied in the future to provide an unbiased prediction using these integrins. In the present study, the association between integrin expression and prognosis, was not analyzed which would have been beneficial to improve the reliability of the proposed integrin mRNA biomarker combination for oral cancer prediction.

Alcohol drinking, betel quid chewing and cigarette smoking are well-known risk factors for oral cancer (47), although non-smoking and betel quid non-chewing individuals may also suffer from oral cancer (48). It is possible that these habits may affect the initiation and progression of oral carcinogenesis, but do not directly affect the overexpression of certain genes in patients with OSCC. For example, no significant differences in the expression levels of the ITGA3, ITGA5, ITGB1 and ITGB6 biomarkers were observed between the OSCC and control tissues for each of these variables.

In conclusion, individual ROC analysis provides a sensitive and specific evaluation of biomarkers for OSCC diagnosis and prognosis. The results of the present study demonstrated that ITGA3 and ITGA5 genes were improved prognostic OSCC biomarkers for all locations compared with ITGB1 and ITGB6 genes, based on individual ROC analysis. However, ITGA3, ITGA5, ITGB1 and ITGB6 genes became more promising OSCC biomarkers when considering specific tumor locations (2 and 3; tongue/mouth part and edentulous ridge, respectively). Additionally, the cumulative ROC analyses indicated that the combination of ITGA3, ITGA5 and ITGB1 genes exhibited the highest AUC values for locations 2/3 and all locations. Therefore, ITGA3, ITGA5, ITGB1 and ITGB6 genes are suggested to be good OSCC biomarkers, and cumulative ROC analysis is hypothesized to provide an improved strategy for cancer biomarker identification.

Acknowledgements

The authors would like to thank Dr Hans-Uwe Dahms; Department of Biomedical Science and Environmental Biology, Kaohsiung Medical University, (Kaohsiung, Taiwan) for English editing.

Funding

The present study was partly supported by funds of the Ministry of Science and Technology, Taiwan (grant no. MOST 104-2320-B-037-013-MY3), the Health and Welfare Surcharge of Tobacco Products, the Ministry of Health and Welfare, Taiwan, Republic of China (grant no. MOHW107-TDU-B-212-114016), ChiMei-KMU Joint Project (grant no. 106CM-KMU-05) and the National Sun Yat-sen University-KMU Joint Research Project (grant no. NSYSU-KMU 107-p001).

Availability of data and materials

The datasets used and/or analyzed in the current study are available from the corresponding author on reasonable request.

Authors' contributions

HWC and CYY contributed significantly in writing the manuscript and data discussion/interpretation. CHC, JHT, and YHK collected the samples and analyzed pathological diagnosis of OSCC. HWC instructed JHT and YTC to extract RNA for qPCR analysis. JYT, YYW, and SSY assisted in clinical data collection and performed statistical analysis. SYL was responsible for study conception and overview. All authors read and approved the final manuscript.

Ethics approval and consent to participate

The present study was approved by the Institutional Review Board at Kaohsiung Medical University (Kaohsiung, Taiwan) (KMUH-IRB-930104), and patients provided written informed consent.

Patient consent for publication

Patient provided written informed consent for the publication of associated data.

Competing interests

The authors declare that they have no competing interests.

References