Application of a non-invasive oral brushing procedure based on bisulfite sequencing of a 13-gene panel to study high-risk OSCC patients

Abstract

BACKGROUND:

A non-invasive sampling procedure for the early detection of Oral Squamous Cell Carcinoma (OSCC) based on DNA methylation analysis of a panel of 13 genes was applied in 4 different OSCC risk-group of patients. Aim of the study is to evaluate the between-group differences and the variables related to the methylation profile of each group.

METHODS:

Oral brushing samples were collected from 54 healthy subjects, 31 Oral Leukoplakia (OL) patients, 18 Oral Lichen Planus (OLP) patients and 26 patients previously treated for OSCC. Each sample was considered positive or negative in relation to a predefined cut-off value.

RESULTS:

None of the samples from 54 healthy subjects were positive, whereas 22/31 OL, 3/18 OLP and 8/26 surgically treated OSCC samples showed positive values with respect to the cut-off. In OL patients, dysplasia was the only variable significantly related to positive values: 10/10 OLs with high-grade dysplasia were positive with respect to 12/21 OLs without dysplasia (Chi 6.039, $p<$ 0.05).

CONCLUSION:

DNA methylation analysis in epithelial cells collected by oral brushing seems to be a promising genetic method to distinguish lesions at high risk of developing OSCC. Larger population studies and an adequate follow-up period are necessary to confirm these preliminary data.

Keywords

Bisulfite sequencing quantitative DNA methylation analysis algorithm oral brushing oral carcinoma oral leukoplakia oral lichen planus

1. Introduction

Like most cancers, the prognosis of OSCC depends to a large extent on how early the cancer is diagnosed and treatment begins. If diagnosed at an stage (I and II), the prognosis is generally from moderate to good: the survival rate for patients with early stage OSCC is around 80% at five years after tumor diagnosis [1]. Unfortunately, even today more than 60% of patients have an OSCC diagnosis in stages III or IV, meaning that their management becomes complex and multidisciplinary.

Screening populations for the early detection of asymptomatic oral carcinoma or precursor lesions is an attractive strategy to reduce the burden of OSCC. The National Screening Committee defines screening as “a process of identifying apparently healthy people who may be at increased risk of a disease or condition” [2]. Screening can be proposed for the whole population or for selected high-risk groups. Screening programs for cancers such as breast, cervix and colorectum have effectively decreased mortality rates and the incidence of these tumors [3, 4]. The potential for OSCC screening exists as there are recognized premalignant phases of the disease (oral potentially malignant disorders, OPMD) during which high-risk subjects could be identified and the detection and treatment of early-stage cancers may reduce mortality [5].

OPMD currently include leukoplakia, erythroplakia, oral lichen planus, oral submucous fibrosis, actinic keratosis and discoid lupus erythematosus. Oral leukoplakia (OL) and oral lichen planus (OLP) are the most common forms of OPMD [6]. OL is defined as a white patch or plaque that cannot be characterized clinically or pathologically as any other disease. Leukoplakia has been postulated to be the clinical expression of genetic alterations within the oral mucosa epithelium whose accumulation can facilitate the development of OSCC [7]. OLP is the consequence of a chronic cell-mediated immune condition of unknown etiology, in which T lymphocytes accumulate beneath the epithelium of the oral mucosa and increase the differentiation rate of the stratified squamous epithelium [8]. OLP inflammatory infiltrate acting chronically on oral mucosa may generate genetic aberrations in keratinocytes which, in case of clonal expansion, may spread in altered fields in which an OSCC could arise [9].

Screening and periodical follow-up may be of primary importance also for OSCC surgical patients. Indeed, patients surgically treated for OSCC can develop a second primary OSCC, with a frequency ranging between 17 and 30% [10, 11]. This rate is greater than other type of tumors and is more often the cause of death [12].

Conventional oral examinations, including visual inspection and palpation, are still today the routine non-invasive methods for the screening and follow-up of oral lesions. However, subtle lesions may pass undetected with oral examination alone, and it is difficult to distinguish between benign, premalignant and malignant lesions. Thomson et al. reported that dysplasia or micro-invasive carcinoma can occur in clinically normal-appearing mucosa [13].

For this reason, several non-invasive techniques have been developed in the last few decades. Technologies that use dyes, autofluorescence, toluidine blue or exfoliative cytology are currently implemented in clinical practice but have not improved early detection rates [2]. At the same time, molecular tests including RNA, microRNA, protein-based panels and hypermethylation are under development but not yet clinically validated [14, 15, 16].

Our research group recently developed a non-invasive method to detect high-risk OSCC lesions starting from oral brushing, quantitatively measuring the DNA methylation level of a panel of 13 genes. In a recent paper, we correctly stratified OSCC and high-grade squamous intraepithelial lesions (HGSIL) from normal donors using a multiclass linear discriminant analysis of a pre-selected 13-gene panel (sensitivity 97.1%, specificity 100%, AUC 0.981) [17].

The purpose of the present study was to apply our non-invasive method in four different groups of patients: a target screening population of healthy individuals as reference controls, two groups of patients with potentially malignant oral cavity disorders (OL and OLP) and a group of patients previously treated for oral cancer during follow-up period. Our aim was to evaluate the between-group differences and the epidemiologic, clinical and histological variables influencing the methylation profile in each group.

2. Material and methods

2.1 Ethics statement

All clinical investigations were conducted according to the principles expressed in the Declaration of Helsinki. The study was approved by the local Ethics Committee (study number 14092, protocol number 899/CE). All information regarding the human material used in this study was managed using anonymous numerical codes. Each participant gave informed consent.

2.2 Study population

This prospective study included 129 oral brushing specimens taken from 4 different groups of patients: a first group of 54 consecutive subjects enrolled in a local bank as a screening population, a second group of 31 patients with a clinicopathological diagnosis of OL, a third group of 18 patients with a clinical and histological diagnosis of OLP and a fourth group of 26 patients previously treated for OSCC. The present cohort included some of the patients analyzed in a previous study [17]. All samples of the present study were consecutively collected from January 1, 2016 to March 31, 2018.

The first group included subjects enrolled during an early detection program for oral cancer organized in a local bank in Bologna, NE Italy. For two months (May and June 2017), a specialist in oral medicine and a maxillofacial surgeon recruited patients older than 40 years. The early detection program included a routine examination of oral mucosa. Potential cases with a high-risk lesion (oral carcinoma or OPMD) were investigated in this group of lesions. Clinicopathological information obtained from this group of patients included age, sex, smoking habits and the presence of lesions in the oral cavity. The second and third groups included patients with potentially malignant lesions referred to the Department of Biomedical and Neuromotor Sciences, University of Bologna, Italy. They consisted of 31 patients with a definitive clinicopathological diagnosis of OL and 18 patients with a clinical and histological diagnosis of OLP.

The definition of OL proposed by the WHO in 2017 [18] was used: ‘A predominantly white plaque of questionable risk having excluded (other) known diseases or disorders that carry no increased risk for cancer’. All patients with a provisional OL diagnosis underwent histological analysis and a definitive diagnosis of OL was made when any etiological cause other than tobacco had been excluded and histopathology had not confirmed any other specific disorder as described by van der Waal et al. [19]. OLP histological diagnosis was based on the presence of irregular acanthosis, degeneration of the basal cell layer of the epithelium, and an inflammatory infiltrate in the upper chorion composed almost exclusively of mature lymphocytes. No dysplastic features were associated with OLP. Clinicopathological information obtained from OL and OLP included age, sex, smoking habits, clinical appearance (homogeneous, non-homogeneous, verrucous in OL patients and reticular or erosive lesions in OLP patients) and the presence of dysplasia.

Patients with high-grade dysplasia in the incisional biopsy were treated surgically and the study included only patients without histological signs of infiltrative oral carcinoma in the excised lesion. Follow-up was performed every three or six months for OL and OLP without dysplasia and the study population included only OL and OLP without neoplastic evolution in the period from January 1, 2016 to January 31, 2018.

Histological examination of potentially malignant lesions was performed blindly by two pathologists (MPF, SA) at the Department of Biomedical and Neuromotor Sciences, “M. Malpighi” Section of Anatomic Pathology, at Bellaria Hospital, University of Bologna, Italy. A multihead microscope discussion was made on discordant cases to obtain a consensus on diagnosis. Lesions were considered dysplastic only in the presence of moderate/severe dysplasia following the criteria described in the WHO/IARC Classification of Tumors, 2017 [18].

Finally, the fourth group comprised 26 consecutive patients who had completed OSCC treatment at least six months prior to brushing collection and with no recurrence of cancer since then. This group included patients with OSCC in remission (OSCCr) who were enrolled during routine follow-up visits after primary OSCC cancer treatment. All 26 patients underwent surgical resection of OSCC in accordance with standard treatment practice [20]. Surgery consisted of composite resections, including excision of the primary tumor with ipsilateral or bilateral neck dissection. Microvascular reconstruction was performed for patients with locally advanced disease. Postoperative radiation therapy was considered an exclusion criterion in the present study.

Follow-up was performed every two weeks for the first two months after surgery and then monthly during the first year after surgery, every three months during the second year after surgery, and finally every six months. A CT scan or MRI was requested every six months during the first three years after surgery and then once a year.

Clinicopathological information obtained from OSCCr patients included age, sex, site, tumor stage according to the TNM classification of the International Union Against Cancer [21], clinically positive cervical lymph node metastasis (LNM) at tumor presentation using criteria defined in van den Brekel et al. [22], and presence or not of a skin graft used for tissue reconstruction after OSCC resection. A surgical margin was considered an exclusion criterion.

2.3 Oral brushing method

A cytobrush was used to collect exfoliated cells from oral mucosa as previously described [17, 23, 24].

Table 1
Genomic coordinates of each of 13 region of interest evaluated in this study. A total of 245 CpG were evaluated in this study (mean for every gene: 19)

Gene	Description	Map	RefSEQ	Position	UCSC h19 coordinates	Amplicon length	Position respect to TSS	Number of interrogated CpGs	Imprinting status
ZAP70	Zeta chain of T-cell receptor associated protein kinase 70	2q11.2	NM_001079	Exon 3	Chr2: 98340750- 98340885	180	10728	20	No
GP1BB	Glycoprotein Ib platelet beta subunit	22q11.21	NM_000407	Exon 1	Chr22: 19710829- 19710960	192	363	18	No
KIF1A	Kinesin family member 1A	2q37.3	NM_001244008	Exon 1	Chr2: 241759586- 241759727	189	$-$ 1	27	No
PARP15	Poly (ADP-ribose) polymerase family member 15	3q21.1	NM_001113523	Exon 1	Chr3: 122296564- 122296725	206	93	19	No
ITGA4	Integrin subunit alpha 4	2q31.3	NM_000885	Exon 2	Chr2: 182322887- 182323053	214	912	14	No
NTM	Neurotrimin	11q25	NM_016522	Exon 1	Chr11: 131781042- 131781185	190	62	15	Yes, maternal, no SNP
MIR193A	microRNA 193a	17q11.2	NR_029710	Promoter	Chr17: 29886860- 29887068	256	$-$ 178	26	No
EPHX3	Epoxide hydrolase 3	19p13.12	NM_024794	Exon 1	Chr19: 15342826- 15343049	223	215	29	No
LINC 00599	Long intergenic non-protein coding RNA 599	8p23.1	NR_029668	Exon 1	Chr8: 9760739- 9760890	199	69	20	No
FLI1	Fli-1 proto-oncogene, ETS transcription factor	11q24.3	NM_001271010	Exon 1	Chr11: 128564020- 128564160	186	187	12	No
MIR296	microRNA 296	20q13.32	NR_029844	Exon 1	Chr20: 57392355- 57392545	238	180	15	Yes, paternal, no SNP
LRRTM1	Leucine rich repeat transmembrane	2p12	NM_178839	Promoter	Chr2: 80531676- 80531807	179	$-$ 431	24	Yes, paternal, No SNP
TERT	Telomerase reverse transcriptase	5p15.33	NM_198253	Intron4-5	Chr5: 1279743- 1279851	109	14976	6	No, SNP at rs10069690

In OPMD lesions and in regenerative areas after OSCC surgical resection all lesion surfaces were gently brushed five times. Brushing cell collection was always performed before incisional biopsy and without any local anesthetic. After brushing, each cytobrush sample was placed in a 2 ml tube containing DNAshield (ZYMO RESEARCH CORP.) and stored at $-$ 20 ${}^{\circ}$ C for DNA/RNA preservation.

2.4 DNA methylation analysis

DNA from exfoliating brush specimens was purified using The MasterPure ${}^{\text{TM}}$ Complete DNA Purification Kit (Epicentre, cod. MC85200). Bisulfite treatment of genomic DNA (200–500 ng) was carried out with the EZDNA Methylation-Lightning ${}^{\text{TM}}$ Kit (Zymo Research, cod. D5031) according to the manufacturer’s protocol. Quantitative DNA methylation analysis was performed as previously described [17]. In brief, well-defined CpG islands of the following 13 genes (see Table 1 for gene coordinates) were amplified by multiplex PCR: ZAP70, ITGA4, KIF1A, PARP15, EPHX3, NTM, LRRTM1, FLI1, MIR193, LINC00599, MIR296, TERT, and GP1BB. The library was prepared using the Nextera ${}^{\text{TM}}$ index kit as previously described [25]. Locus-specific bisulfite amplicon libraries were generated with tagged primers using Phusion U DNA polymerase (ThermoFisher, cod. F555L).

Sequencing was conducted on the MiSEQ (Illumina, cod. 15027617) according to the manufacturer’s protocol. FASTQ output files already trimmed for the multiplex identifier were processed for quality control and converted into FASTA format in a Galaxy Project environment [26]. To evaluate the methylation ratio of each CpG, we loaded FASTA files into the bisulfite sequencing pattern analysis tool (BSPAT – http://cbc.case.edu/BSPAT/index.jsp) [27]. In parallel, the same evaluation was done using BWA meth in a galaxy project environment (Europe) followed by MethylDackel tool, and finally using Kismeth [28].

A ROC curve analysis applied for discriminating OSCC and HGSIL from normal donors in our previous study identified a threshold of 1.0615547 as the best value for sensitivity and specificity (AUC $=$ 0.981) [17]. Following the same algorithm in the present study, a specific score was calculated for each sample. Values exceeding the threshold of 1.0615547 were considered positive.

2.5 Statistical analysis

Each sample was considered either a numeric or dicotomic variable (pos/neg) in the statistical analysis. One-way ANOVA analysis with multiple range test and Chi square analysis were used to evaluate any between-group (healthy subjects, OL group, OLP group and OSCCr group) significant differences and clinical and histological variables influencing the methylation profile of each group. $P$ values $<$ 0.05 were considered statistically significant in all analyses.

3. Results

3.1 Study population

3.1.1 Group 1 (healthy subjects)

Fifty-four subjects enrolled during an early detection program for oral cancer (27 males and 27 females) had a median age of 46.33 $\pm$ 10.58. None of the subjects examined had a suspect OSCC lesion or a suspect potentially malignant lesion.

3.1.2 Group 2 (OL)

14/31 (45.2%) OL patients were males and 17/31 (54.8%) OL patients were females (median age 63.35 $\pm$ 14.15). 24/31 (77.4%) OL patients were no smokers while 7/31 (22.6%) were smokers. 7/31 (22.6%) patients showed a homogeneous OL while the remaining 24/31 (77.4%) showed an OL with a dysomogeneous clinical aspect. Finally, histological examination revealed the presence of 10/31 (32.3%) OL with high grade dysplasia and 21/31 (67.7%) with absence of it. All 10 lesions with presence of high-grade dysplasia were surgically excised. One patient with a diagnosis of OL without presence of dysplasia developed an OSCC during follow up period of the present study.

3.1.3 Group 3 (OLP)

Six out of 18 (33.33%) OLP patients were males and 12/18 (66.67%) were females (median age 61.36 $\pm$ 12.06). Fourteen out of 18 (77.78%) OLP patients were non-smokers and the remaining 4/18 (22.22%) were smokers. Clinical features revealed reticular lesions in 11/18 (61.11%) patients while other seven patients were clinically characterized by an erosive form (38.89%). None of the 18 OLP (0%) showed histological architectural alterations.

Figure 1.

Columns obtained using the cut-off values showed a significant between-group difference (Chi square analysis Chi 53.23; $p<$ 0.05).

3.1.4 Group 4 (OSCCr)

13/26 (50%) OSCCr patients were males and the remaining 13 were females (50%) (median age 65.77 $\pm$ 12.81). 4/26 (15.4%) patients of group 4 were smokers and other 22/26 (84.6%) were non-smokers. 12/26 (46.2%) patients were diagnosed as a T1-2 at time of neoplastic presentation while 14/26 (53.8%) showed a tumor size classified as T3-4. 20/26 (76.9%) showed an N negative at time of tumor diagnosis while 6/26 (23.1%) patients showed a loco-regional infiltration with N positive. In 12/26 cases (46.2%) a skin graft was used to reconstruct OSCC defect while in the remaining 14/26 (53.8%) a skin graft reconstruction was not necessary. 4/26 (15.4%) OSCCr patients developed a second neoplastic event during the follow up period of the present study. Three patients developed a secondary tumor located in oral cavity (1 Local Recurrence and 2 Second Primary Tumours) and 1 patient developed a Lymph-node metastasis.

3.2 DNA methylation analysis: score calculation and statistical analysis

A mean score of $-$ 1.17127 $\pm$ 0.210271 was calculated in the group of 54 healthy subjects. None of the subjects exceeded the threshold value and were all considered negative.

A mean score value of 2.02 $\pm$ 0.26 was calculated in the group of 31 OL patients. Twenty-two out of 31 (70.9%) OL patients exceeded the threshold value and were considered positive. Presence of high grade dysplasia resulted the only variable statistically related to the presence of a positive score in OL. Indeed, 10/10 (100%) OL patients with high grade dysplasia exceeded to threshold value with respect to 12/21 (57.1%) patients whose lesions lacked of dysplasia at the histology (Chi 6.039, $p<$ 0.05). The non-dysplastic OL patient who developed an OSCC during follow up period presented a positive score.

A mean score value of 0.23041 $\pm$ 0.3642 was calculated in the group of 18 OLP patients. Three out of 18 (16.67%) OLP patients exceeded the threshold value and were considered positive.

A mean score value of 0.71 $\pm$ 0.28 was calculated in the group of 26 OSCCr patients. Eight out of 26 (30.7%) patients exceeded the threshold value and were considered positive. Three out of 4 patients who developed a second neoplastic manifestation presented a positive score. Interestingly, 3 patients who experienced a secondary tumor in oral cavity showed positive scores, whereas the patient who developed a Lymph-node metastasis scored negatively.

A significant difference between groups was detected after Chi square analysis (Chi 53.23, $p<$ 0.05, Fig. 1) and one-way ANOVA analysis (F $=$ 34.66, $p<$ 0.05, Fig. 2).

Figure 2.

Scatterplot obtained using the scores calculated from the algorithm showed a significant between-group difference (One way ANOVA test $F=$ 34.66, $p<$ 0.05).

The post hoc analysis, performed to compare the effect size in the different groups, underlined that OL patients showed higher values with respect to other groups. On the other hand, healthy subjects showed significant lower values with respect to other groups (Table 2).

Table 2

Multiple range test showed significant higher values in the generated score between OL group and other groups. Healthy subjects showed significant lower values with respect to other groups. Finally, No significant differences in the generated score between OLP group and OSCCr group was found

Method: 95,0 percent LSD
Pathology	Count	Mean	Homogeneous groups
Healthy subjects	54	$-$ 1.17127	A
OLP	18	0.23041	B
OSCCr	26	0.707816	B
OL	31	2.02174	C
Contrast	Sig.	Difference	$+$ / $-$ Limits
Healthy subjects vs. OL	${}^{*}$	$-$ 3.19301	0.635305
Healthy subjects vs. OLP	${}^{*}$	$-$ 1.40168	0.767332
Healthy subjects vs. OSCCr	${}^{*}$	$-$ 1.87908	0.672995
OL vs. OLP	${}^{*}$	1.79133	0.835471
OL vs. OSCCr	${}^{*}$	1.31392	0.749756
OLP vs. OSCCr		$-$ 0.477406	0.864477

None of clinico-pathological variables examined in other groups resulted significantly related to the presence of a positive score calculation. In particular, no differences in the mean score value were found between smokers and no smokers in the group of healthy subjects.

Table 3 summarizes the results of all variables analyzed.

Table 3

In Table 3 all variables analyzed were summarized

Groups	Clinicopathological variables		Positive score		Score meanvalue
Healthy subjects	Age ( $<$ 50 vs. $>$ 50)	$<$ 50	0/31		$-$ 1.28	$P=$ 0.097
		$>$ 50	0/23		$-$ 1.01
	Sex (f vs. m)	Female	0/27		$-$ 1.15	$P=$ 0.88
		Male	0/27		$-$ 1.18
	Smoking habits (no	No	0/35		$-$ 1.10	$P=$ 0.26
	vs. yes)	Yes	0/19		$-$ 1.3
OL patients	Age ( $<$ 50 vs. $>$ 50)	$<$ 50	6/6	Chi 3.04	2.99	$F=$ 2.38
		$>$ 50	16/25	$P=$ 0.07	1.79	$P=$ 0.13
	Sex (f vs. m)	Female	11/17	Chi 0.72	1.98	$F=$ 0.02
		Male	11/14	$P=$ 0.39	2.07	$P=$ 0.89
	Smoking habits (no	No	19/24	Chi 3.47	2.3	$F=$ 2.9
	vs. yes)	Yes	3/7	$P=$ 0.06	1.05	$P=$ 0.09
	Clinical feature	Homogeneous	3/7	Chi 3.47	0.876818	$F=$ 3.86
		Dysomogeneous	19/24	$P=$ 0.06	2.30877	$P=$ 0.0595
	Dysplasia	No dysplasia	12/21	Chi 6.04	1.62992	$F=$ 3.5
		High grade dysplasia	10/10	$P=$ 0.01 ${}^{*}$	2.779	$P=$ 0.07
OLP patients	Age ( $<$ 50 vs. $>$ 50)	$<$ 50	0/3	Chi 0.72	$-$ 0.096291	$F=$ 0.27
		$>$ 50	3/15	$P=$ 0.4	0.29575	$P=$ 0.61
	Sex (f vs. m)	Female	2/12	Chi 0.00	0.256055	$F=$ 0.02
		Male	1/6	$P=$ 1	0.179121	$P=$ 0.9
	Smoking habits (no	No	3/14	Chi 1.03	0.348644	$F=$ 0.64
	vs. yes)	Yes	0/4	$P=$ 0.31	$-$ 0.183408	$P=$ 0.44
	OLP-OLL	OLP	1/15	Chi 0.72	0.115794	$F=$ 0.87
		OLL	1/3	$P=$ 0.4	0.803493	$P=$ 0.37
	Clinical feature	Reticular	1/11	Chi 1.17	0.177011	$F=$ 0.06
		Erosive	2/7	$P=$ 0.28	0.314323	$P=$ 0.83
OSCCr patients	Age ( $<$ 60 vs. $>$ 60)	$<$ 60	2/11	Chi 1.420	$-$ 0.27	$F=$ 4.31
		$>$ 60	6/15	$P=$ 0.23	1.42	$P=$ 0.05
	Sex (f vs. m)	Female	6/13	Chi 2.89	1.1	$F=$ 0.84
		Male	2/13	$P=$ 0.09	0.32	$P=$ 0.37
	Smoking habits (no	No	7/22	Chi 0.7	0.92	$F=$ 1.4
	vs. yes)	Yes	1/4	$P=$ 0.79	$-$ 0.46	$P=$ 0.25
	T stage	T1-2	2/12	Chi 2.09	0.21	$F=$ 0.81
		T3-4	6/14	$P=$ 0.15	0.38	$P=$ 0.38
	N stage	N neg	6/20	Chi 0.02	1.38	$F=$ 1.38
		N pos	2/6	$P=$ 0.88	0.25	$P=$ 0.25
	Skin graft	Skin graft NO	3/10	Chi 0.01	0.59	$F=$ 0.34
		Skin graft YES	5/16	$P=$ 0.95	0.75	$P=$ 0.57

3.3 Principal component analysis

Using the principal component analysis with the highest distribution of data (PC1) as the x-axis and the second highest principal component (PC2) as the y-axis, the data are distributed as evenly across the plot as possible while maintaining the distance between points as a proxy for how similar each point is to the other. The graph shows that OL cases with dysplasia (blue) are located in the left center of the plot, while healthy subjects (violet) are clustered in a well-defined and restricted area. OSCCr, OLP and OL without dysplasia have a less defined distribution (Fig. 3).

Figure 3.

Principal component analysis (PCA): Unit variance scaling is applied to rows; SVD with imputation is used to calculate principal components. X and Y axis show principal component 1 and principal component 2 that explain 44.5% and 16.3% of the total variance, respectively. Prediction ellipses are such that with probability 0.95, a new observation from the same group will fall inside the ellipse. $N=$ 129 data points.

4. Discussion

At time of writing, oral biopsy with histological assessment remains the gold standard for the diagnosis of OSCC or OPMD. Nevertheless, biopsy is an invasive surgical approach that can create discomfort and may be refused by the patient, especially in “benign-looking lesions”. In addition it is hardly applicable as a routinary and consistent diagnostic tool in the follow up of patients at risk of developing OSCC [29].

Conventional oral examination based on visual and tactile assessment is the non-invasive method most frequently guiding identification and follow up of a suspect lesion in the oral cavity. Early diagnosis of OSCC or a true OPMD requires a high clinical index of suspicion stemming from both experience and knowledge. The difficulty in differentiating premalignant and malignant lesions from benign or inflammatory lesions with similar clinical features reflects the well-known limits of conventional oral examination [13].

Consequently, non-invasive detection techniques are needed to identify and monitor high-risk lesions in the oral cavity. Recent findings indicated that quantitative DNA methylation analysis of a specific set of genes may be an attractive strategy to detect oral cancer lesions even at the early stage [25, 30], and location of core regions and the density of methylation are required for gene silencing [31]. Our group and other authors recently proposed analyzing the methylation status of a panel of different genes starting from non-invasive sampling methods (saliva and/or brushing procedure) [23, 32, 33, 34, 35, 36]. Our previous study reported the development of a novel assay to detect early OSCC and HGSIL from oral brushing specimens using bisulfite next generation sequencing [17]. Using quantitative DNA methylation analysis of ZAP70, ITGA4, KIF1A, PARP15, EPHX3, NTM, LRRTM1, FLI1, MIR193, LINC00599, MIR296, TERT and GP1BB, we clearly discriminated OSCC and HGSIL from healthy donors. In particular, OSCC and HGSIL showed hypermethylation of ZAP70, ITGA4, KIF1A, PARP15, EPHX3, NTM, LRRTM1, FLI1, MIR193, LINC00599 and hypomethylation of TERT, MIR296 and GP1BB.

In the same study, using a linear discriminant analysis, we calculated a score that weighted the best CpGs from the 13 selected genes investigated. The power of this new assay to discriminate oral cancer from healthy donors reached an optimal level of accuracy (AUC: 0.981): in particular 28 out of 29 OSCC (96.6%) and six out of six HGSIL (100%) specimens exceeded the threshold value, whereas no specimens from the 65 healthy donors exceeded the threshold.

The high sensitivity and specificity of the method stimulated us to apply the procedure to oral mucosa at high risk of developing OSCC, i.e. mucosa with OPMDs and mucosa regenerated after OSCC removal, to elucidate the intrinsic diagnostic and prognostic potential of our assay.

The results of the present study showed that the highest level of positive samples occurred in the majority of patients with Oral Leukoplakia (22/31, 70.9%). OL is the most commonly diagnosed preneoplastic lesion and shows a rate of malignant transformation between 17% and 24% [7, 37]. In many cancers, gene silencing by promoter methylation seems to be an early event in carcinogenesis and may occur even earlier than structural inactivation of genes through mutations or deletions. It implies that non pathologic tissue adjacent to tumors and OPMDs can have aberrant methylation patterns in candidate genes even in absence of histological or strictly genetic abnormalities [16]. This could be true also for oral carcinogenesis.

In addition, positive scores resulted highly related to the presence of high-grade dysplasia, as all ten patients with high grade dysplasia exceeded the threshold.

Dysplasia is a very strong parameter when assessing the risk of malignant transformation. In fact, it is widely accepted that the more severe the degree of dysplasia is encountered the greater is the likelihood of progression to malignancy. The association found in the present study between presence of high-grade dysplasia and aberrant methylation patterns seems to confirm the accuracy of this non-invasive assay in detecting lesions considered at risk.

On the opposite, in the group of OLs with no dysplasia a great variability of results was observed, as an altered methylation profile identified with a positive score was found in more than half of samples (12/21, 57.1%). This data seems to confirm that identification of an altered methylation pattern is an early event in oral carcinogenesis. The presence of a positive score in the only patient with non-dysplastic OL that developed an OSCC is promising, giving to our new algorithm higher predictive power than conventional histology.

Otherwise, larger follow up studies with an adequate follow up are needed to disclose the predictive power of our method and to identify one or more genes of our 13-gene panel able to distinguish OL with high-grade dysplasia and OL with absence of dysplasia with a higher accuracy. However, it would be noteworthy in the clinical application of our procedure if in the future none of the OLs with negative score evolved into OSCC. If confirmed, this procedure based on DNA methylation may represent a first level diagnosis to identify patients who would mostly benefit from a more intensive diagnostic approach (i.e. restricted follow up, multiple incisional biopsies).

A second interesting finding concerns the data obtained in the group of 26 surgically treated OSCC patients. In this group, the oral brushing specimen was collected in the surgical area after OSCC treatment and samples were enrolled during routine follow-up visits after primary OSCC cancer treatment, at least six months after OSCC surgery. Eight of 26 (30.7%) OSCCr patients showed a positive score indicating an altered epigenetic pattern in oral brushing specimens from normal-appearing buccal mucosa, probably related to a field cancerization effect detectable after surgical resection in close proximity to the index tumor. On the other hand, we cannot exclude that these alterations are related to the regenerative process itself, such as active cell proliferation or inflammation. During the period of this study 4 of 26 patients developed a second neoplastic manifestation. A positive score was detected only in 3/4 (75%) OSCCr patients who developed a second neoplastic manifestation, but all three positive cases developed a secondary tumor in oral cavity; on the contrary the only negative patient with a second neoplastic event developed an extra-oral lymph-node metastasis distant from site of oral brushing cell collection. These data, albeit taken with caution due to the relatively small size of the cohort, indicate that it should be advisable to perform the epigenetic test periodically after surgical resection, in order to early detect the presence of a second local event. Further studies are needed to elucidate if an intraoral brushing sampling may have or not a predictive value for extraoral neoplastic spread.

Our population study identified 3/18 (16.6%) OLP exceeding the threshold, a significantly lower number with respect to OL samples. The WHO also included OLP among the risk conditions for malignant transformation, even if it is difficult to evaluate the real risk of malignant transformation in OLP patients [6]. Studies describing OLP malignant transformation reported highly variable rates ranging from 0.5% to 2% [38]. As a result, the malignant nature of OLP is so confusing and controversial that recent studies have even considered OLP a benign condition [39]. The three OLP patients with a positive score in the present study are now undergoing a strict follow-up to obtain more information on this point. If these patients will not manifest any malignant lesion, they may be considered as false positives, probably due to alterations in some genes involved in lichenoid inflammation processes. A further study with a larger population of OLP patients is needed to investigate the potential role of one or more genes of our 13-gene panel in the wide spectrum of lichenoid manifestations.

We also applied our non-invasive procedure in a small group of 54 subjects enrolled during an early detection program for Oral Cancer organized in a local bank. None of patients of this small group showed the presence of a malignant or premalignant lesion during routine examination and developed an OSCC during follow up period; and this is in agreement considering the incidence of OSCC in general population (8.4–12.1 new cases per 100000 inhabitants per year [40]). All 54 specimens collected from healthy subjects were calculated as negative by our method and variables as smoking habits, sex and age didn’t significantly influence the methylation profile in this group of subjects confirming the high specificity previously identified.

Admittedly the oral brushing procedure here adopted has some limitations. We used oral brushing as non-invasive method to obtain DNA for quantitative methylation analysis. Brush cytology can collect cells from external mucosa of the lesion and/or from different sites of the oral cavity in the same patient. Although it is considered an easy to perform method for collecting cells, a correct training is necessary to avoid inappropriate sampling procedure, leading to the acquisition of insufficient numbers of specific cells from the lesion. Furthermore, the 13-gene DNA methylation analysis is a time-consuming procedure: 7–10 days are usually necessary for quantitative methylation analysis and score calculation. Anyway, this time is not excessive for a procedure proposed to measure oral cancer risk for suspicious lesions or surveillance for high-risk OSCC patients.

5. Conclusions

This study applied a non-invasive oral brushing procedure based on DNA methylation analysis of a pre-selected set of 13 genes in different groups of patients at risk of developing OSCC. None of the healthy individuals exceeded the threshold value and positive values were detected only in groups of patients with a recognized risk of developing an OSCC (OL, OLP and OSCCr patients). The group of patients with OL showed a significantly higher positive score with respect to the other groups, with a significant difference between dysplastic and non-dysplastic lesions. High scores were also found in a third of patients previously treated for OSCC and patients who develop a neoplastic lesion in oral cavity (1 OL patient and 3 OSCCr patients) showed positive values. If preliminary results will be confirmed in a larger cohort of patients with an adequate follow up period this procedure may be proposed not in replacement of oral biopsy for definitive diagnosis but as an easy-to-perform and reliable measure of oral cancer risk to alert primary care providers, dentists, oral medicine specialists and other frontline screeners.

Footnotes

Acknowledgments

The english in this document has been checked by a professional editor native speaker of English. The authors thank Stesi Kavaja and Laura Felicetti (DDS of Department of Biomedical and Neuromotor Sciences, Section of Oral Sciences, University of Bologna) for support in sample collection. We would like to thank all patients and healthy subjects who participated in the study. This study was supported by an academic grant (FARB-FFBO124539) from the University of Bologna.

Conflict of interest

As a possible conflict of interest, Luca Morandi, Davide B Gissi, and Achille Tarsitano submitted a patent (the applicant is the University of Bologna) in November 2016 to the National Institute of Industrial Property; however, we believe that this is a natural step of translational research (bench-to-bedside) and guarantee that the scientific results are true. The remaining authors declare that they have no competing interests.

References

Rusthoven

K.E.

Raben

Song

J.I.

Kane

Altoos

T.A.

and Chen

, Survival and patterns of relapse in patients with oral tongue cancer, J. Oral Maxillofac. Surg 68 (2010), 584–589. doi: 10.1016/j.joms.2009.03.056.

Brocklehurst

Kujan

Glenny

A.-M.

Oliver

Sloan

Ogden

and Shepherd

, Screening programmes for the early detection and prevention of oral cancer, Cochrane Database Syst Rev (2010), CD004150. doi: 10.1002/14651858.CD004150.pub3.

Hewitson

Glasziou

Irwig

Towler

and Watson

, Screening for colorectal cancer using the faecal occult blood test, Hemoccult, Cochrane Database Syst Rev (2007), CD001216. doi: 10.1002/14651858.CD001216.pub2.

Gøtzsche

P.C.

and Nielsen

, Screening for breast cancer with mammography, Cochrane Database Syst Rev (2011), CD001877. doi: 10.1002/14651858.CD001877.pub4.

Ganly

Patel

and Shah

, Early stage squamous cell cancer of the oral tongue-clinicopathologic features affecting outcome, Cancer 118 (2012), 101–111. doi: 10.1002/cncr.26229.

Warnakulasuriya

Johnson

N.W.

and van der Waal

, Nomenclature and classification of potentially malignant disorders of the oral mucosa, J. Oral Pathol. Med 36 (2007), 575–580. doi: 10.1111/j.1600-0714.2007.00582.x.

Warnakulasuriya

and Ariyawardana

, Malignant transformation of oral leukoplakia: a systematic review of observational studies, J. Oral Pathol. Med 45 (2016), 155–166. doi: 10.1111/jop.12339.

Roopashree

M.R.

Gondhalekar

R.V.

Shashikanth

M.C.

George

Thippeswamy

S.H.

and Shukla

, Pathogenesis of oral lichen planus – a review, J. Oral Pathol. Med 39 (2010), 729–734. doi: 10.1111/j.1600-0714.2010.00946.x.

Mignogna

M.D.

Fedele

Lo Russo

Mignogna

de Rosa

and Porter

S.R.

, Field cancerization in oral lichen planus, Eur J Surg Oncol 33 (2007), 383–389. doi: 10.1016/j.ejso.2006.09.028.

10.

Braakhuis

B.J.M.

Tabor

M.P.

Leemans

C.R.

van der Waal

Snow

G.B.

and Brakenhoff

R.H.

, Second primary tumors and field cancerization in oral and oropharyngeal cancer: molecular techniques provide new insights and definitions, Head Neck 24 (2002), 198–206.

11.

Montebugnoli

Leonardi

Morandi

Farnedi

Gissi

D.B.

Marchetti

Tarsitano

Balbi

Gentile

Cocchi

and Foschini

M.P.

, Genetic relationship between multiple squamous cell carcinomas arising in the oral cavity, Head Neck 36 (2014), 94–100. doi: 10.1002/hed.23259.

12.

Lippman

S.M.

and Hong

W.K.

, Second malignant tumors in head and neck squamous cell carcinoma: the overshadowing threat for patients with early-stage disease, Int. J. Radiat. Oncol. Biol. Phys 17 (1989), 691–694.

13.

Pentenero

Marino

Tempia Valenta

Navone

and Gandolfo

, Microbiopsy a novel sampling technique to early detect dysplastic/malignant alterations in oral mucosal lesions: practicability by general dentists, J. Oral Pathol. Med 43 (2014), 435–440. doi: 10.1111/jop.12161.

14.

Arellano

Boontheung

Wang

Zhou

Jiang

Elashoff

Wei

Loo

J.A.

and Wong

D.T.

, Salivary proteomics for oral cancer biomarker discovery, Clin. Cancer Res 14 (2008), 6246–6252. doi: 10.1158/1078-0432.CCR-07-5037.

15.

Carvalho

A.L.

Jeronimo

Kim

M.M.

Henrique

Zhang

Hoque

M.O.

Chang

Brait

Nayak

C.S.

Jiang

W.-W.

Claybourne

Tokumaru

Lee

Goldenberg

Garrett-Mayer

Goodman

Moon

Koch

Westra

W.H.

Sidransky

and Califano

J.A.

, Evaluation of promoter hypermethylation detection in body fluids as a screening/diagnosis tool for head and neck squamous cell carcinoma, Clin. Cancer Res 14 (2008), 97–107. doi: 10.1158/1078-0432.CCR-07-0722.

16.

Cheng

Y.-S.L.

Rees

and Wright

, A review of research on salivary biomarkers for oral cancer detection, Clin Transl Med 3 (2014), 3. doi: 10.1186/2001-1326-3-3.

17.

Morandi

Gissi

Tarsitano

Asioli

Gabusi

Marchetti

Montebugnoli

and Foschini

M.P.

, CpG location and methylation level are crucial factors for the early detection of oral squamous cell carcinoma in brushing samples using bisulfite sequencing of a 13-gene panel, Clin Epigenetics 9 (2017), 85. doi: 10.1186/s13148-017-0386-7.

18.

El-Naggar

A.K.

, What is new in the world health organization 2017 histopathology classification? Curr Treat Options Oncol 18 (2017), 43. doi: 10.1007/s11864-017-0469-x.

19.

van der Waal

, Oral leukoplakia, the ongoing discussion on definition and terminology, Med Oral Patol Oral Cir Bucal 20 (2015), e685–692.

20.

Kademani

Bell

R.B.

Bagheri

Holmgren

Dierks

Potter

and Homer

, Prognostic factors in intraoral squamous cell carcinoma: the influence of histologic grade, J. Oral Maxillofac. Surg 63 (2005), 1599–1605. doi: 10.1016/j.joms.2005.07.011.

21.

AJCC cancer staging manual MB Amin, SB Edge – 2017 – Springer, n.d.

22.

van den Brekel

M.W.

Stel

H.V.

Castelijns

J.A.

Nauta

J.J.

van der Waal

Valk

Meyer

C.J.

and Snow

G.B.

, Cervical lymph node metastasis: assessment of radiologic criteria, Radiology 177 (1990), 379–384. doi: 10.1148/radiology.177.2.2217772.

23.

Morandi

Gissi

Tarsitano

Asioli

Monti

Del Corso

Marchetti

Montebugnoli

and Foschini

M.P.

, DNA methylation analysis by bisulfite next-generation sequencing for early detection of oral squamous cell carcinoma and high-grade squamous intraepithelial lesion from oral brushing, J Craniomaxillofac Surg 43 (2015), 1494–1500. doi: 10.1016/j.jcms.2015.07.028.

24.

Gissi

D.B.

Morandi

Gabusi

Tarsitano

Marchetti

Cura

Palmieri

Montebugnoli

Asioli

Foschini

M.P.

and Scapoli

, A noninvasive test for MicroRNA expression in oral squamous cell carcinoma, Int J Mol Sci 19 (2018). doi: 10.3390/ijms19061789.

25.

Sandoval

and Esteller

, Cancer epigenomics: beyond genomics, Curr. Opin. Genet. Dev 22 (2012), 50–55. doi: 10.1016/j.gde.2012.02.008.

26.

Afgan

Baker

Batut

van den Beek

Bouvier

Cech

Chilton

Clements

Coraor

Grüning

B.A.

Guerler

Hillman-Jackson

Hiltemann

Jalili

Rasche

Soranzo

Goecks

Taylor

Nekrutenko

and Blankenberg

, The Galaxy platform for accessible, reproducible and collaborative biomedical analyses: 2018 update, Nucleic Acids Res 46 (2018), W537–W544. doi: 10.1093/nar/gky379.

27.

Ting

A.H.

and Li

, BSPAT: a fast online tool for DNA methylation co-occurrence pattern analysis based on high-throughput bisulfite sequencing data, BMC Bioinformatics 16 (2015), 220. doi: 10.1186/s12859-015-0649-2.

28.

Gruntman

Slotkin

R.K.

Roeder

Martienssen

R.A.

and Sachidanandam

, Kismeth: analyzer of plant methylation states through bisulfite sequencing, BMC Bioinformatics 9 (2008), 371. doi: 10.1186/1471-2105-9-371.

29.

Yang

E.C.

Tan

M.T.

Schwarz

R.A.

Richards-Kortum

R.R.

Gillenwater

A.M.

and Vigneswaran

, Noninvasive diagnostic adjuncts for the evaluation of potentially premalignant oral epithelial lesions: current limitations and future directions, Oral Surg Oral Med Oral Pathol Oral Radiol 125 (2018), 670–681. doi: 10.1016/j.oooo.2018.02.020.

30.

Y.-F.

Hsiao

Y.-H.

Lai

Y.-H.

Chen

Y.-C.

Chen

Y.-J.

Chou

J.-L.

Chan

M.W.Y.

Lin

Y.-H.

Tsou

Y.-A.

Tsai

M.-H.

and Tai

C.-K.

, DNA methylation profiles and biomarkers of oral squamous cell carcinoma, Epigenetics 10 (2015), 229–236. doi: 10.1080/15592294.2015.1006506.

31.

van Vlodrop

I.J.H.

Niessen

H.E.C.

Derks

Baldewijns

M.M.L.L.

van Criekinge

Herman

J.G.

and van Engeland

, Analysis of promoter CpG island hypermethylation in cancer: location, location, location! Clin. Cancer Res 17 (2011), 4225–4231. doi: 10.1158/1078-0432.CCR-10-3394.

32.

Demokan

Chang

Chuang

Mydlarz

W.K.

Kaur

Huang

Khan

Ostrow

K.L.

Brait

Hoque

M.O.

Liegeois

N.J.

Sidransky

Koch

and Califano

J.A.

, KIF1A and EDNRB are differentially methylated in primary HNSCC and salivary rinses, Int. J. Cancer 127 (2010), 2351–2359. doi: 10.1002/ijc.25248.

33.

Langevin

S.M.

Stone

R.A.

Bunker

C.H.

Grandis

J.R.

Sobol

R.W.

and Taioli

, MicroRNA-137 promoter methylation in oral rinses from patients with squamous cell carcinoma of the head and neck is associated with gender and body mass index, Carcinogenesis 31 (2010), 864–870. doi: 10.1093/carcin/bgq051.

34.

Pattani

K.M.

Zhang

Demokan

Glazer

Loyo

Goodman

Sidransky

Bermudez

Jean-Charles

McCaffrey

Padhya

Phelan

Spivakovsky

Bowne

H.Y.

Goldberg

J.D.

Rolnitzky

Robbins

Kerr

A.R.

Sirois

and Califano

J.A.

, Endothelin receptor type B gene promoter hypermethylation in salivary rinses is independently associated with risk of oral cavity cancer and premalignancy, Cancer Prev Res (Phila) 3 (2010), 1093–1103. doi: 10.1158/1940-6207.CAPR-10-0115.

35.

Nagata

Hamada

Yamada

Yokoyama

Kitamoto

Kanmura

Nomura

Kamikawa

Yonezawa

and Sugihara

, Aberrant DNA methylation of tumor-related genes in oral rinse: a noninvasive method for detection of oral squamous cell carcinoma, Cancer 118 (2012), 4298–4308. doi: 10.1002/cncr.27417.

36.

Schussel

Zhou

X.C.

Zhang

Pattani

Bermudez

Jean-Charles

McCaffrey

Padhya

Phelan

Spivakovsky

Brait

Bowne

H.Y.

Goldberg

J.D.

Rolnitzky

Robbins

Kerr

A.R.

Sirois

and Califano

J.A.

, EDNRB and DCC salivary rinse hypermethylation has a similar performance as expert clinical examination in discrimination of oral cancer/dysplasia versus benign lesions, Clin. Cancer Res 19 (2013), 3268–3275. doi: 10.1158/1078-0432.CCR-12-3496.

37.

Gissi

D.B.

Gabusi

Servidio

Cervellati

and Montebugnoli

, Predictive role of p53 protein as a single marker or associated with ki67 antigen in oral leukoplakia: a retrospective longitudinal study, Open Dent J 9 (2015), 41–45. doi: 10.2174/1874210601509010041.

38.

Accurso

B.T.

Warner

B.M.

Knobloch

T.J.

Weghorst

C.M.

Shumway

B.S.

Allen

C.M.

and Kalmar

J.R.

, Allelic imbalance in oral lichen planus and assessment of its classification as a premalignant condition, Oral Surg Oral Med Oral Pathol Oral Radiol Endod 112 (2011), 359–366. doi: 10.1016/j.tripleo.2011.03.042.

39.

Zhang

Michelsen

Cheng

Zeng

Priddy

and Rosin

M.P.

, Molecular analysis of oral lichen planus. A premalignant lesion? Am. J. Pathol 151 (1997), 323–327.

40.

Warnakulasuriya

, Global epidemiology of oral and oropharyngeal cancer, Oral Oncol 45 (2009), 309–316. doi: 10.1016/j.oraloncology.2008.06.002.

Application of a non-invasive oral brushing procedure based on bisulfite sequencing of a 13-gene panel to study high-risk OSCC patients

Abstract

BACKGROUND:

METHODS:

RESULTS:

CONCLUSION:

Keywords

1. Introduction

2. Material and methods

2.1 Ethics statement

2.2 Study population

2.3 Oral brushing method

Table 1 Genomic coordinates of each of 13 region of interest evaluated in this study. A total of 245 CpG were evaluated in this study (mean for every gene: 19)

2.5 Statistical analysis

3. Results

3.1 Study population

3.1.1 Group 1 (healthy subjects)

3.1.2 Group 2 (OL)

3.1.3 Group 3 (OLP)

3.2 DNA methylation analysis: score calculation and statistical analysis

5. Conclusions

Footnotes

Acknowledgments

Conflict of interest

References

Table 1
Genomic coordinates of each of 13 region of interest evaluated in this study. A total of 245 CpG were evaluated in this study (mean for every gene: 19)