ORIGINAL PAPER

Genetic gains and losses in oral squamous cell carcinoma:impact on clinical management

Ilda Patrícia Ribeiro & Francisco Marques & Francisco Caramelo & João Pereira &

Miguel Patrício & Hugo Prazeres & José Ferrão & Maria José Julião &

Miguel Castelo-Branco & Joana Barbosa de Melo & Isabel Poiares Baptista &

Isabel Marques Carreira

Accepted: 5 November 2013 /Published online: 19 December 2013# International Society for Cellular Oncology 2013

AbstractPurpose The identification of genetic markers associated withoral cancer is considered essential to improve the diagnosis,prognosis, early tumor and relapse detection and, ultimately,to delineate individualized therapeutic approaches. Here, weaimed at identifying such markers.Methods Multiplex Ligation-dependent Probe Amplification(MLPA) analyses encompassing 133 cancer-related geneswere performed on a panel of primary oral tumor samplesand its corresponding resection margins (macroscopicallytumor-free tissue) allowing, in both types of tissue, the detec-tion of a wide arrange of copy number imbalances on varioushuman chromosomes.Results We found that in tumor tissue, from the 133 cancer-related genes included in this study, those that most frequentlyexhibited copy number gains were located on chromosomal

arms 3q, 6p, 8q, 11q, 16p, 16q, 17p, 17q and 19q, whereasthose most frequently exhibiting copy number losses werelocated on chromosomal arms 2q, 3p, 4q, 5q, 8p, 9p, 11qand 18q. Several imbalances were highlighted, i.e., losses ofERBB4 , CTNNB1 , NFKB1 , IL2 , IL12B , TUSC3, CDKN2A ,CASP1 , and gains of MME , BCL6 , VEGF, PTK2 , PTP4A3 ,RNF139 , CCND1 , FGF3 , CTTN , MVP, CDH1 , BRCA1 ,CDKN2D, BAX, as well as exon 4 of TP53 . Comparisonsbetween tumor and matched macroscopically tumor-freetissues allowed us to build a logistic regression model topredict the tissue type (benign versus malignant). In thismodel, the TUSC3 gene showed statistical significance,indicating that loss of this gene may serve as a goodindicator of malignancy.Conclusions Our results point towards relevance of the abovementioned cancer-related genes as putative genetic markers

Electronic supplementary material The online version of this article(doi:10.1007/s13402-013-0161-5) contains supplementary material,which is available to authorized users.

I. P. Ribeiro : J. Ferrão : J. B. de Melo : I. M. Carreira (*)Cytogenetics and Genomics Laboratory, Faculty of Medicine,University of Coimbra, Polo Ciências da Saúde,3000-354 Coimbra, Portugale-mail: .citogenetica@fmed.uc.pt

I. M. Carreirae-mail: icarreira@fmed.uc.pt

I. P. Ribeiro : F. Marques : J. B. de Melo : I. P. Baptista :I. M. CarreiraCIMAGO - Center of Investigation on Environment Genetics andOncobiology - Faculty of Medicine, University of Coimbra,3000-354 Coimbra, Portugal

F. Marques : I. P. BaptistaDepartment of Dentistry, Faculty of Medicine, University ofCoimbra, 3000-075 Coimbra, Portugal

F. MarquesStomatology Unit, Coimbra Hospital and University Centre, CHUC,EPE, 3000-075 Coimbra, Portugal

F. Caramelo : J. Pereira :M. Patrício :M. Castelo-BrancoLaboratory of Biostatistics and Medical Informatics, IBILI - Facultyof Medicine, University of Coimbra, 3000-354 Coimbra, Portugal

H. PrazeresMolecular Pathology Laboratory, Portuguese Institute of Oncologyof Coimbra FG, EPE, 3000-075 Coimbra, Portugal

M. J. JuliãoDepartment of Pathology, Coimbra Hospital and University Centre,CHUC, EPE, 3000-075 Coimbra, Portugal

Cell Oncol. (2014) 37:29–39DOI 10.1007/s13402-013-0161-5

for oral cancer. For practical clinical purposes, these geneticmarkers should be validated in additional studies.

Keywords Oral squamous cell carcinoma . Genetic profile .

Chromosomal imbalances . Copy number losses and gains

1 Introduction

Oral cavity tumors constitute a subgroup of head and necktumors that rank sixth in prevalence, with an annual incidenceof almost 600.000 cases worldwide [1]. The most commonhistological subtype is oral squamous cell carcinoma (OSCC).Tumors of the oral cavity exhibit a complex etiology involv-ing multiple environmental, toxic, and viral factors. In addi-tion to tobacco and alcohol consumption, human papillomavirus (HPV) infection is a well-known risk factor for OSCC[2]. In spite of advances that have been made in diagnostictechnologies and treatment modalities, these tumors are stilldiagnosed at relatively late stages and, consequently, no majorimprovements in survival rates have been made. Patients witha positive OSCC diagnosis undergoing primary treatmentshow recurrence rates ranging from 25–45 % [3–8].

Early detection is considered the gold key for decreasingmorbidity and mortality rates, as well as for reducinghealthcare costs. OSCC development results from the accu-mulation of both genetic and epigenetic changes, and studieshave been reported aimed at identifying aberrantly expressedgenes that can be used in the classification, diagnosis andprognosis of OSCC, including the prediction of treatmentoutcome [9,10]. Although copy number imbalances have beenreported to occur in almost all chromosomes, it appears thatsome chromosomal regions are recurrently affected in thesetumors [11]. Overall, however, oral cancer displays a vastgenetic and biologic heterogeneity, and the most challengingtask is to establish the clinical relevance of each molecularsubgroup associated with specific histopathological features.Therefore, establishing correlations between molecular dataand disease phenotypes appears to be crucial to (i) confirm thehistological type and the stage of the tumor and (ii) predictmore accurately the patient’s outcome. Due to the currentlylimited clinical and pathological capability of identifying pa-tients at high-risk of treatment failure, better biomarkers forprognosis are urgently needed. Ultimately, it will be impera-tive to take into account the genetic profile of each individualpatient in order to be able to delineate personalized therapeuticstrategies. In the present study, we have established the geneticprofiles of 35 OSCC samples and correlated the results ob-tained with its corresponding clinicopathological characteris-tics. The putative relevance of the most frequent geneticchanges encountered, including their applicability in routineclinical practice, is discussed.

2 Materials and methods

2.1 Tumors and control samples

The present study was conducted on 35 primary oral tumorsamples with 28 corresponding resection margins (macro-scopically tumor-free tissue) from 35 patients. These sampleswere obtained between 2010 and 2012 from the MaxillofacialSurgery and Stomatology Unit of the Coimbra Hospital andUniversity Centre, CHUC, EPE, Portugal. All patients weresubmitted to surgery and the histopathologic diagnoses of themirror sections of the samples were performed by two differ-ent pathologists. Hematoxylin and eosin staining was used toevaluate the tumor content in each specimen. In our cohort, allsamples contained at least 50 % tumor cells. Diagnosis andstaging were performed according to the American JointCommittee on Cancer TNM staging system [12] for OSCCs.All patients provided informed consent in accordance with theregulations in the Declaration of Helsinki. The study wasapproved by the Ethics committee of the Faculty of Medicineof the University of Coimbra. Detailed characteristics of ourOSCC cohort are listed in Table 1. As controls gingival tissuesfrom healthy donors subjected to “wisdom teeth” removalwere included. DNAs from patient and control samples wereextracted using a High Pure PCR Template Preparation Kit(Roche GmbH, Mannheim, Germany) according to the man-ufacturer’s instructions, and quantified using a Nanodrop1000 Spectrophotometer (Thermo Scientific, USA).

2.2 Multiplex ligation-dependent probe amplification(MLPA)

MLPAwas performed using four tumor-specific MLPA probepanels. Overall, these four panels (P005, P006, P007 andP014; MRC-Holland, Amsterdam, The Netherlands) included154 probes targeting 133 different genes located on all humanautosomes (supplementary Table 1). The P014 panel wasexclusively designed for chromosome 8, allowing a morecomprehensive study of this chromosome as compared tothe other ones. Details of the probe sequences, gene loci andchromosomal locations can be found at: www.mrc-holland.com. All MLPA reactions were performed according to theprotocol described by Schouten et al. [13]. Briefly, DNAsamples (5 μl) were heated at 98 °C for 10 min. After theaddition of the probe mix, samples were heated for 1 min at95 °C and then incubated for 16 h at 60 °C. Ligation of theannealed oligonucleotide probes was performed for 15 min at54 °C in buffer containing Ligase-65. After inactivating theligase by heating at 98 °C for 5 min, multiplex PCR wascarried out using FAM-labeled primers, dNTPs and SALSApolymerase. PCRs were performed for 35 cycles of 30 s at95 °C, 30 s at 60 °C and 1 min at 72 °C. All the reactions werecarried out in a thermal cycler equipped with a heat lid (ABI

30 I.P. Ribeiro et al.

2720, Applied Biosystems, Foster City, CA, USA). Finally,the PCR products were heat denatured and analyzed using aGene Scan ABI PRISM 3130 capillary electrophoresis system(Applied Biosystems, Foster City, CA, USA). Three normalcontrols and a negative control (without DNA) were includedin each MLPA assay. The results are displayed as ratiosbetween references and experimental samples. For eachMLPA probe we determined specific cut-off values for gainand loss, using the values limiting the 95 % confidenceinterval (CI) as determined on non-cancer samples. Anumerical gain was scored when the ratio was higher than1.2 and a numerical loss was defined when the ratio waslower than 0.8.

2.3 HPV typing

All tumor tissue samples were analyzed for HPV infection asdescribed by Nobre et al. [14]. Briefly, PCR was performedusing established general consensus and degenerate primersets, i.e., GP5+/GP6+ and MY09/MY11, which were de-signed to amplify a fragment of the L1 gene of mucosatropicHPVs. For genotyping we performed Sanger sequencing ofintra-primer segments within the amplified fragments in orderto determine the specific types of HPV present in the samples.In addition to DNA sequencing, we complemented our anal-yses byDNAmicroarray hybridizations, using HPVCLART2arrays (Genomica), to address cases that showed infectionwith multiple HPV genotypes.

2.4 Statistical analysis

The statistical analysis was carried out using the statisticalsoftware package IBM SPSS Statistics for Windows, Version20.0. Armonk, NY: IBMCorp. The significance level adoptedwas p =0.05. From detailed descriptive analyses and chi-square tests corrected for multiple comparisons (Bonferronicorrection), it was possible to reduce the number of genes withstatistical meaning for distinguishing tumor tissue from mac-roscopically tumor-free tissue, selecting only those that weresignificantly imbalanced between the two groups. We thenperformed a logistic regression (forward: conditional) modelusing these genes (thirteen in total), in order to assess theirusefulness as predictors of the tissue type.

3 Results

3.1 Genetic profiles of tumor tissues and macroscopicallytumor-free tissues

All oral tumor samples included in this study were analyzedby MLPA in order to establish their genetic profiles. The sexchromosomes were excluded from the analyses since thecontrol and tumor samples were not gender-matched. Weobserved genetic alterations in all 35 tumor samples analyzedand, in addition, in 23 of the 28 macroscopically tumor-freetissue samples recovered from surgical margins (Fig. 1a). Thenumbers of alterations detected in both tissues were, however,very different (Fig. 1a, b). As expected, we detected morecopy number imbalances in the tumor tissues than in themacroscopically tumor-free tissues. Besides this, the distribu-tion of the imbalances in terms of losses and gains by chro-mosome was very consistent for some chromosomes, i.e., 3pand 8p showed mostly losses, whereas 3q and 8q frequentlyshowed gains; chromosomes 4 and 5 only showed losses forthe genes analyzed. Additionally, on chromosomes 19 and 20we encountered more frequently gains than losses for the

Table 1 Patient and tumor characteristics

Patients (n =35)

Mean age, years (range) 61.5 (37–84)

Sex

Male 30

Female 5

Smoking (cigarettes/day)

≥ 20 21

< 20 3

None 11

Alcohol

Yes 4

None 4

Not recorded 27

Stage

I and II 14

III and IV 21

Site

Tongue 13

Floor of the mouth 12

Buccal mucosa 4

Retromolar trigone 4

Gingival 2

Pathological margin status

Positive 2

Negative 33

Treatment

Surgery + RT 9

CT + Surgery 3

Surgery + RT + CT 7

Surgery only 16

Clinical outcome

Alive 22

Death from the disease 12

Dead from the other cause 1

RT radiation therapy, CT Chemotherapy

Genetic gains and losses in oral squamous cell carcinoma 31

genes analyzed. In the tumor tissues of these 35 patients, wefound that from the 133 genes analyzed, those with themost frequent gains were localized on chromosomal arms3q, 6p, 8q, 11q, 16p, 16q, 17p, 17q and 19q, whereas thosewith the most frequent losses were localized on chromo-somal arms 2q, 3p, 4q, 5q, 8p, 9p, 11q and 18q (Fig. 1b).

In the macroscopically tumor-free tissues, the most frequentgains were localized on chromosomal arms 6p, 16p, 17p,17q and 19q, whereas the most frequent losses were local-ized on chromosomal arms 3p and 9p (Fig. 1b). In contrast,we found that the genes tested on chromosomes 10 and 15did not exhibit imbalances (gains or losses) in more than

Fig. 1 Genetic imbalances (gains and losses) in 35 OSCC patientsdetected using four MLPA probe mixes. Losses of genetic material arerepresented in red, gains in blue. a The results above the black linecorrespond to macroscopically tumor-free tissue and the results belowthis line correspond to tumor tissue. Each line represents one patientand each pixel in this line corresponds to one gene. Gray representsgenes without alteration, and each shade of gray shows the localiza-tion of the genes on each specific chromosome. From left to right,

the genes are ordered by chromosome, from the short arm to thelong arm. Each white pixel on macroscopically tumor-free tissuemeans no information, since we did not have macroscopically tu-mor-free tissue for all patients analyzed. b Picture showing thepercentage of imbalances by chromosome in tumor tissue and inmacroscopically tumor-free tissue for each gene analyzed, excludingthe imbalances detected in HPV-positive patients. Each arrow repre-sents one gene

32 I.P. Ribeiro et al.

two patients. Overall, we found that the number of samples thatshowed losses of genetic material was lower than that showinggains. Chromosome 8 showed most gains and losses in thelargest number of patients (Fig. 1b). It is important note herethat in this study we used one probe panel specific for chromo-some 8 with probes for 30 genes mapped on this chromosome,which allowed a more comprehensive assessment of this chro-mosome compared to the other ones. We also found that nosingle gene was altered in all patients. Figure 2 illustrates themost commonly altered genes on the 12 aforementioned chro-mosomes in tumor tissues, as well as inmacroscopically tumor-free tissues. Thus, from the 133 genes analyzed (Fig. 2), thefollowing were the ones showing frequent losses in tumorsamples: ERBB4 (2q33.3-q34), CTNNB1 (3p21), NFKB1(4q24), IL2 (4q26-q27), IL12B (5q31.1-q33.1), TUSC3(8p22), CDKN2A (9p21) and CASP1 (11q23). It is also im-portant to note that we detected homozygous deletions for boththeCDKN2A andCDKN2B genes in one patient. Additionally,the following genes: MME (3q25.2), BCL6 (3q27), Hs.570518 (3q28), VEGF (6p21.1), PTK2 (8q24.3), PTP4A3(8q24.3), RNF139 (8q24), CCND1 , FGF3 , CTTN (11q13),MVP (16p11.2), CDH1 (16q22.1), exon 4 of TP53 (17p13.1),BRCA1 (17q21),CDKN2D (19p13) andBAX (19q13.3-q13.4)frequently showed gains in the tumor samples. In the macro-scopically tumor-free tissues the genes that showed the mostfrequent losses were: PIK3CA (3q26.3) and CDKN2A (9p21),whereas VEGF (6p21.1), KCNK9 (8q24.3), MVP (16p11.2),exon 4 of TP53 (17p13.1), BRCA1 (17q21), CRK (17p13.3),CDKN2D (19p13) and BAX (19q13.3-q13.4), showed themost frequent gains.

3.2 Genetic imbalances predicting clinicopathologicalfeatures

Next, we generated a logistic regression model to predict thetype of tissue (tumor versus macroscopically tumor-free)

using the DLGAP2, TUSC3, EXT1, RNF139, MYC, DDEF1,PTK2, PTP4A3 and RECOL4 genes on chromosome 8, andthe CCND1, FGF3, CTTN and BIRC2 genes on chromosome11 as predictors. The final model (−2LL=24.004; Cox andSnell R2=0.630; Nagelkerke R2=0.843) included the TUSC3,PTK2 and CCDN1 genes, but only the TUSC3 gene showedstatistical significance (p =0.041). The accuracy of this modelwas 93.7 %, compared to 55.6 % if the prediction had beenrandom. In this model, the probability of being tumor tissuewas 27-fold higher when the TUSC3 gene showed loss versusthis gene being normal (OR=27.000 with CI 95 % [2.091;348.661]). If we only take the TUSC3 gene into account tomake this regression model, the accuracy drops to 68.3 %(−2LL=69.852; Cox and Snell R2=0.233; Nagelkerke R2=0.312). Nonetheless, in this simplified model the value of thisgene remains statistically significant (p =0.017), with an OR=21.316 for losses with CI 95 % [2.590; 175.398].

With respect to other clinicopathological features, includ-ing tumor stage, development of metastasis and tobacco con-sumption, we found similar patterns of copy number lossesand gains across the genome in both stage I + II and stage III +IV, as well as in the presence or absence of metastases, and inthe smokers and non-smokers groups (Fig. 3). No single geneshowed statistical significance, and we were unable to genet-ically differentiate patients belonging to stages I or II fromthose belonging to stages III or IV. The same lack of statisticalsignificance was found for the presence or absence of metas-tases, and for the consumption or non-consumption of tobac-co. We did, however, observe differences in losses of geneticmaterial between smokers and non-smokers, i.e., only thesmokers showed losses at 3p (MLH1 ) and 11q (ATM).

3.3 HPV infections in OSCC samples

Among the 35 tumor samples analyzed, two were found to beHPV-positive (data not shown). One of these samples showed

Fig. 2 Radial heatmap of thegenes frequently altered in 12most commonly affectedchromosomes for tumor tissueand for macroscopically tumor-free tissue based on the use of fourMLPA probe sets. Each linerepresents one patient. Red linesrepresent losses of geneticmaterial and blue lines representgains

Genetic gains and losses in oral squamous cell carcinoma 33

a HPV type belonging to a low-risk class, i.e., HPV type 42,whereas the other sample showed a HPV type belonging to ahigh-risk class, i.e., HPV type 31. The low-risk HPV patientdid not exhibit any other risk factors, such as tobacco use oralcohol consumption.

4 Discussion

Cancer is considered to be a disease of the genome, and toresult from the sequential acquisition of DNA alterations bysomatic cells. Besides being useful for early diagnostics, thesealterations may also serve as specific targets for therapy. Assuch, they provide a window of hope and promise. Since theidentity of the most relevant oral cancer-related genes is stillunknown, we set out to identify at least some of them throughgenomic profiling.

4.1 Genomic profiling of oral cancer

Until now, no single gene alteration has been identified that isexclusive for oral cancer. After analyzing in detail the mostcommonly altered chromosomes in our cohort, we found thatwe could highlight specific genes that frequently show copynumber gains or losses. Thus, on chromosome 2 we observedfrequent losses of the ERBB4 gene (2q33.3-q34) only intumor tissue. In the past, in breast cancer decreased ERBB4protein expression has been correlated with increased recur-rence rates [15]. This gene should, therefore be taken intoaccount for OSCC in order to redefine its predictive value forrecurrence risk and, consequently, for choosing additionaltreatment options. On chromosome 3, we found that mostfrequent losses and gains occurred at 3p and 3q, respectively.In a wide range of patient samples we observed copy numbergains at 3q, highlighting the genes MME (3q25.2), BLC6(3q27), Hs.570518 (3q28) and IL12A (3q25.33) as beingputatively related to oral carcinogenesis. Indeed, previouslyFreier et al. [16] detected frequent DNA copy number gains at3q, and emphasized the possibility of several candidate proto-oncogenes being located within the region 3q25-qter. Intrigu-ingly, PIK3CA (3q26.3) showed both gains and losses at thesame frequencies in tumor tissues and, additionally, in mac-roscopically tumor-free tissues this gene was the only one onthis chromosome that showed alterations in a substantial num-ber of patients (8.5 %). In another study on head and neckcancer, the aforementioned gene had already been pointed outas a strong candidate oncogene [17]. Losses observed on 3phighlighted several putative tumor suppressor genes, such asCTNNB1 (3p21), MLH1 (3p21.3) and VHL (3p25.3). Previ-ously, losses of CTNNB1 (3p21) have significantly beenassociated with precursor lesions showing progression to-wards laryngeal carcinoma [18]. It will be crucial to furtherelucidate the role of these genes in oral tumor initiation and to

assess their role in tumor progression during clinical follow-up of the patients. Previously chromosome 3p has been de-scribed as being frequently involved in loss of heterozygosity(LOH) in OSCC. So, other tumor suppressor genes locatedwithin this region, such as FHIT, may also be relevant for thedevelopment of this carcinoma [19]. On chromosome 4 weonly observed copy number losses, highlighting the genesNFKB1 (4q24) and IL2 (4q26-q27). Loss of the NFKB1 genewas also observed in the macroscopically tumor-free tissue ofone patient. Deletions associated with this chromosome wereinitially considered to be relatively rare as compared to otherchromosomal aberrations. However, Pershouse et al. [20]revealed that 92 % of head and neck tumors showed deletionsinvolving chromosome 4 with the highest frequency of loss inband q25 [20,21]. Until now, no tumor suppressor gene withinthis region has firmly been established as being important fororal tumor development. On chromosome 5, inactivation ofgenes located at 5q21-22 is thought to be common and, thus,they may be associated with the initiation or progression ofOSCC [22]. Besides that, we might speculate about the puta-tive importance of the IL12B (5q31.1–q33.1), IL4 (5q31.1)and RAD17 (5q13) genes, which frequently showed losses inour cohort. On chromosome 6 we observed a preferential gainof the VEGF (6p21.1) gene, both in tumor tissue and inmacroscopically tumor-free tissue. Previously, the expressionof VEGF and its receptors has been found to increase duringtumor growth [23], thus turning them into potential targets forcancer therapy. In our cohort, copy number alterations onchromosome 8 were observed in the largest number of patientsamples. We showed that in two samples all the genes ana-lyzed on 8q exhibited copy number gains, and that threesamples showed losses of all genes analyzed on 8p. Theseresults are in line with other studies and are compatible withthe formation of isochromosomes 8q [24]. We find, however,that isochromosomes do not represent the main rearrange-ments that occur in this chromosome, since our data, as wellas data reported by others, strongly suggest gains and lossesthat do not match with such a scenario. We found that gains in8q were more frequent than losses in 8p. Similar to our results,Lin et al. [25] identified gain of 8q as the most prevalentchromosomal anomaly in these tumors. Band 8q24, whichharbors the MYC gene, is considered to be most important.Garnis et al. [26] raised the possibility of additional oncogenesnear the MYC gene to be relevant for the progression of oralcancer. The same authors [27] observed amplifications in band8q22 and postulated that the LRP12 gene, which maps in this

�Fig. 3 Percentages of imbalances by chromosome in 35 OSCC tumors.Each arrow represents one gene. Losses of genetic material is representedin red , gains in blue . a Genetic profiles of tumors in stage I or II, and instage III or IV. b Genetic profiles of tumors that developed metastases,and tumor without metastases. c Genetic profiles of tumors in smokersand non-smokers

34 I.P. Ribeiro et al.

Genetic gains and losses in oral squamous cell carcinoma 35

region, could be implicated in oral carcinogenesis. Our studyshowed that the distal region of chromosome 8 is most fre-quently altered, i.e., the PTK2 (8q24.3) gene followed by thePTP4A3 (8q24.3) gene. The MYC and LRP12 genes exhib-ited less frequent gains (Fig. 2). Gains in PTK2 copy numbershave also been reported by others [28–30], suggesting itsrelevance for oral carcinogenesis. Thus, in our cohort it seemsthat two events are prevalent: amplification of the 8q24 regionand isochromosome 8q formation with a concomitant loss of8p. These events were not always seen in the same tumors.Regarding the losses at 8p, our study highlights the geneTUSC3 (8p22) as being putatively important to oral carcino-genesis. This gene has already been suggested as an importanttarget of genomic rearrangements in epithelial cancers [31].On chromosome 9 we observed preferential copy numberlosses at 9p21, where the genes CDKN2A and CDKN2B arelocated. The CDKN2A and CDKN2B genes are positioned intandem, spanning a region of approximately 80 kb, withCDKN2B located 25 kb centromeric to CDKN2A [32]. TheCDKN2A gene is considered to be the major tumor suppressorgene that is targeted in a wide variety of human cancers [33].In our study, we identifiedmore frequently losses ofCDKN2Athan of CDKN2B , both in tumor tissues and in macroscopi-cally tumor-free tissues, supporting the idea thatCDKN2A hasa more important role in oral carcinogenesis than CDKN2B.This genomic imbalance has been pointed out as the mostcommon of all genetic changes occurring in the early progres-sion of oral tumors [34]. Concerning the second chromosomemost commonly altered in terms of gain of genetic material,i.e., chromosome 11, we found that band 11q13 was mostfrequently increased (57 %). Similarly, others have reportedamplifications of band 11q13 in about 30–45 % of head andneck cancers [35,36]. Initially, the CCND1 gene was consid-ered to be the most important 11q13 target gene [37–40].More recently, however, other candidate genes have beenconsidered as possible cofactors of CCND1, i.e., FGF3 andCTTN [37]. In our study we observed co-amplification ofthese three genes in 20 patient samples, which corroboratesthe findings described in the literature and provides strongevidence for the importance of these genes in oral carcinogen-esis. However, copy number gain of other genes located in thisregion, such as the TMEM16A gene, may also contribute tooral carcinogenesis [41]. We also identified losses in 11q,mostly the CASP1 (11q23) gene. Additionally, two othergenes appear to be interesting, i.e., ATM (11q22-q23) andBIRC2 (11q22). The last one showed copy number lossesand gains in seven and eight patients, respectively. As op-posed to the gains in the 11q13 region, where some geneswere already identified and correlated to clinical outcome,such information is as yet scarce on the losses observed in11q. Parikh et al. [36] proposed that haploinsufficiency orcopy number loss of the ATM gene may contribute to defectsin DNA damage responses and reduced sensitivity to ionizing

radiation, which consequently may lead to tumor progression.Based on our data, we can raise the hypothesis that not onlythe ATM gene may be important but also, and perhaps evenmore so, the CASP1 gene, since its encoded cysteine-asparticacid protease 1 is known to play a central role in apoptosis.Thus, loss of this gene may contribute to a decrease in apo-ptotic signaling. Interestingly, this may open options for pro-apoptotic drugs as a valid choice for treatment [42]. Onchromosome 16 we identified mostly gains of the MVP(16p11.2) andCDH1 (16q22.1) genes.MVP/vaults have beenassociated with chemoresistance in primary tumors and vari-ous tumor cell lines. Therefore,MVP is frequently consideredas a negative prognostic factor for response to chemotherapy,as well as disease-free survival and/or overall survival [43].The CDH1 gene, which encodes E-cadherin, is one of themost important genes regulating cell-cell adhesion in epithe-lial tissues [44]. This gene is considered to be an invasion-suppressor gene and loss of function of its encoded protein hasbeen correlated with increased invasiveness and metastaticpotential of tumors [44]. On chromosome 17 we detectedfrequent copy number gains of the BRCA1 (17q21), TP53(17p13.1) and CRK (17p13.3) genes. Previously, Hardissonet al. [45] detected chromosome 17 anomalies using fluores-cence in situ hybridization (FISH) analyses in pharynx andlarynx carcinomas. As TP53 is well-known for its role asguardian of the genome, loss was expected for this region.Additionally, besides deletions, TP53 gene mutations andprotein inactivation were also reported [46]. On chromosome18 we observed gene losses on its q-arm. Some of thesegenes may specifically be important for oral carcinogen-esis, i.e., CDH2 (18q11.2), BCL2 (18q21.3) and DCC(18q21.3). In head and neck cancer loss of 18q is com-monly observed, and the putative importance of the DCCgene for these tumors has already been highlighted [29].On chromosome 19 we identified copy number gains inall genes analyzed. The BAX (19q13.3–q13.4) andCDKN2D (19p13) genes showed gains both in tumortissues and in macroscopically tumor-free tissues. Amplifica-tions of the 19q13 region in oral and esophageal carcinomashave already been described [47,48]. Unexpectedly, wefound that some of the genes that have previously beendescribed as tumor suppressor genes were amplified in ourstudy. This observation could be explained by the fact thatmassive DNA rearrangements occurred randomly in someof the chromosomes. Some of these rearranged regionsmay harbor both tumor suppressor genes and dominantlyacting oncogenes which, ultimately, may have led to again of these regions. A simpler explanation could be thatthese tumor suppressor genes are non-functional (e.g. dueto hypermethylation) and, thus may have been amplifiedas a passenger event in samples showing genetic instability.Clearly, further studies are required in order to correctly inter-pret these rearrangements.

36 I.P. Ribeiro et al.

4.2 Importance of evaluating surgical margins

It is worth noting that surgical margins in two of the patientsincluded in our study were histologically evaluated as tumorpositive. Surgical margins represent macroscopically tumor-free tissues. We detected genetic imbalances in both histolog-ically positive and negative margins. Delineating the exactarea of excision is a great challenge for physicians, since thepresence of tumor cells within or close to a surgical marginmay be indicative for a risk of relapse which, in turn, affectsadditional treatment options [49]. Unfortunately, relapses alsooccur in patients with histologically tumor-free margins aftersurgery. Therefore, our understanding of the transition fromnormal mucosa to potentially malignant oral mucosa, andfrom that to tumor needs to be explored in more detail. Severalquestions need to be addressed, such as how many geneticevents need to occur in a cell that clonally expands andspreads to normal epithelium in order to create a field oftransformed epithelium? Is there a common early geneticevent that drives the development of multiple tumors or re-currences? If at the onset of carcinogenesis the cells do carrygenetic alterations but at the histological level seem to benormal, what happens genetically, in quantitative and qualita-tive terms, to make the distinction between a histologicallynormal and a histologically malignant appearance? The mo-lecular understanding of this transition is crucial for the de-velopment and implementation of biomarkers in routine diag-nostics. The analysis of macroscopically tumor-free tissuesfrom surgical margins made us aware of the fact that thesesurgical margins frequently show genetic imbalances similarto those encountered in the tumor from the same patient. Suchfindings may be indicative for an increased risk of relapses. Inlight of that, a rigorous follow-up of these patients seemsmandatory, as well as the development of a non-invasiveway to perform this surveillance.

4.3 Predictionmodels correlating genetic profiles with clinicalfeatures

Despite the fact that genomic profiling can be extremelyuseful for distinguishing different tumor sub-types, most prob-ably because of the relatively small number of patients en-rolled, we could not make a distinction between oral tumorswith different clinicopathologic features on basis on theirgenomic profiles. We found that the genomic profiles oftumors in stages I or II and tumors in stages III or IV werevery similar. The same was found for patients that exhibitedmetastases and those who did not, as well as for smokers andnon-smokers. We did find, however, that smokers exhibitedlosses at 3p (MLH1) and 11q (ATM ) (Fig. 3c), suggesting theoccurrence of specific imbalances that could be characteristicfor smokers. Such data might be important for classifyingOSCCpatients in different subgroupswith different prognostics

as well as different therapeutic responses. We anticipate thatwith a larger cohort it may be possible to molecularly identifysuch OSCC subgroups.

Califano et al. [50] were the first to describe a progressionmodel for OSCC. In this model, losses at chromosomal re-gions 3p, 9p and 17p were considered early events in thecarcinogenic process. In the present study, comparisons be-tween tumor and matched macroscopically tumor-free tissuesallowed us to build a logistic regression model to predict thetwo types of tissue. By applying this model, the TUSC3 geneturned out to be the only one that reached statistical signifi-cance, which may be indicative for its relevance in the devel-opment of oral tumors. Thus, the TUSC3 gene may play a rolein the transition from normal oral mucosa to potentially ma-lignant oral mucosa. Precise and adequate prediction modelsare essential to determine the patients eligibility for clinicaltrials and to predict the disease outcome, as well as to selectindividualized therapies for each case.

4.4 Role of HPV typing

In this study, we only identified two HPV-positive patientsamples. This result is not surprising taking into account thatin the group of head and neck cancers HPV infection has beenreported most particularly in association with oropharynxcarcinoma, where its incidence may in some cases reach upto 60 % [51,52]. It has been reported that HPV-positive andHPV-negative tumors may exhibit distinct clinicopathologicaland molecular features [53]. In our study it was impossible toassess whether HPV-positive cases represent a distinct groupwith a genetic profile different from HPV-negative cases(Fig. 1a). It is relevant to note here that the mean age of ourcohort was 61.5 years, whereas the incidence of oral tumorshas increased mostly among younger people. Perhaps in thesecases HPV vaccination could be an option.

In spite of continuous technological progress, our under-standing of oral tumors is still limited. It is, therefore, peremp-tory to identify genes that may serve as good candidates forfurther studies, in order to validate them as biomarkers and totranslate their application into routine clinical practice. Ourcurrent results not only reinforce previous reports, but alsorevealed novel imbalances in chromosomes 2, 3, 4, 5, 6, 8, 9,11, 16, 17, 18 and 19, with a putative impact in terms ofclinical management. Selection of the most frequently alteredgenes may be instrumental for the development of biomarkersdistinguishing between different susceptibilities for relapsesand, possibly, different chemotherapeutic agents. In order todistinguish tumor tissue from tumor-free tissue, the TUSC3gene may serve as a bona fide biomarker. In the future thelogistic regression model presented here could help cliniciansto optimize the clinical management of OSCC patients byimproving the estimation of the risk of relapses, the survivalrates and, ultimately, the prognosis.

Genetic gains and losses in oral squamous cell carcinoma 37

Acknowledgments The authors are grateful to Dr. Artur Ferreira,Director of the Maxillofacial Surgery Unit from Coimbra Hospitaland University Centre, for his contribution in the collection of thesamples. This work was supported in part by CIMAGO (Center ofInvestigation on Environment Genetics and Oncobiology - Faculty ofMedicine, University of Coimbra).

Conflict of interest The authors declare that they have no conflict ofinterest.

References

1. A. Jemal, R. Siegel, E. Ward, Y. Hao, J. Xu, M.J. Thun, Cancerstatistics, 2009. CA Cancer J. Clin. 59(4), 225–249 (2009)

2. S. Ambatipudi, M. Gerstung, R. Gowda, P. Pai, A.M. Borges, A.A.Schaffer, N. Beerenwinkel, M.B. Mahimkar, Genomic profiling ofadvanced-stage oral cancers reveals chromosome 11q alterations asmarkers of poor clinical outcome. PLoS One 6(2), e17250 (2011)

3. L.P. Kowalski, Results of salvage treatment of the neck in patientswith oral cancer. Arch. Otolaryngol. Head Neck Surg. 128(1), 58–62(2002)

4. S.A. Liu, Y.K. Wong, J.C. Lin, C.K. Poon, K.C. Tung, W.C. Tsai,Impact of recurrence interval on survival of oral cavity squamous cellcarcinoma patients after local relapse. Otolaryngol. Head Neck Surg.136(1), 112–118 (2007)

5. A. Nathanson, K. Ågren, A. Biörklund,M.G. Lind, L. Andréason,M.Anniko, A. Freijd, L. Lejdeborn, S. Kinman, A. Kumlien, A. Köling,J. Lindström, A. Novik, J. Olofsson, K.-L. Pettersson, Evaluation ofsome prognostic factors in small squamous cell carcinoma of themobile tongue: a multicenter study in Sweden. Head Neck 11(5),387–392 (1989)

6. W.R. Rate, P. Garrett, R. Hamaker, M. Singer, N. Pugh, D. Ross, R.Haerr, G. Charles, Intraoperative radiation therapy for recurrent headand neck cancer. Cancer 67(11), 2738–2740 (1991)

7. W.Wang, S. Han, Z. Yao, X. Li, P. Huang, M. Zhang, Y. Chen, J. He,A study of epidemiologic and recurrence factors of oral cancer. J.Oral Maxillofac. Surg. (2011)

8. L.Y. Wong, W.I. Wei, L.K. Lam, A.P. Yuen, Salvage of recurrenthead and neck squamous cell carcinoma after primary curative sur-gery. Head Neck 25(11), 953–959 (2003)

9. B.J. Braakhuis, R.H. Brakenhoff, C.R. Leemans, Second fieldtumors: a new opportunity for cancer prevention? Oncologist10(7), 493–500 (2005)

10. D. Weiss, C. Stockmann, K. Schrodter, C. Rudack, Protein expres-sion and promoter methylation of the candidate biomarker TCF21 inhead and neck squamous cell carcinoma. Cell Oncol (Dordr) 36(3),213–224 (2013)

11. M.A. Guervos, C.A. Marcos, M. Hermsen, A.S. Nuno, C. Suarez,J.L. Llorente, Deletions of N33, STK11 and TP53 are involved in thedevelopment of lymph node metastasis in larynx and pharynx carci-nomas. Cell. Oncol. 29(4), 327–334 (2007)

12. C. Wittekind, F.L. Greene, R.V.P. Hutter, M. Klimpfinger, L.H.Sobin, TNM atlas. Illustrated Guide to the TNM/pTNMClassification of Malignant Tumours, 5th edn. (UICC), I.U.A.C.,(Springer-Verlag, Berlin-Heidelberg-New York, 2003)

13. J.P. Schouten, C.J. McElgunn, R. Waaijer, D. Zwijnenburg, F.Diepvens, G. Pals, Relative quantification of 40 nucleic acid sequencesby multiplex ligation-dependent probe amplification. Nucleic AcidsRes. 30(12), e57 (2002)

14. R.J. Nobre, E. Cruz, O. Real, L.P. de Almeida, T.C. Martins,Characterization of common and rare human papillomaviruses inPortuguese women by the polymerase chain reaction, restriction

fragment length polymorphism and sequencing. J. Med. Virol.82(6), 1024–1032 (2010)

15. N.L. Barnes, S. Khavari, G.P. Boland, A. Cramer, W.F. Knox, N.J.Bundred, Absence of HER4 expression predicts recurrence of ductalcarcinoma in situ of the breast. Clin. Cancer Res. 11(6), 2163–2168(2005)

16. K. Freier, C. Hofele, K. Knoepfle,M. Gross, F. Devens, G. Dyckhoff,P. Plinkert, P. Lichter, C. Herold-Mende, Cytogenetic characteriza-tion of head and neck squamous cell carcinoma cell lines as modelsystems for the functional analyses of tumor-associated genes. J. OralPathol. Med. 39(5), 382–389 (2010)

17. B. Singh, S.K. Gogineni, P.G. Sacks, A.R. Shaha, J.P. Shah, A.Stoffel, P.H. Rao, Molecular cytogenetic characterization of headand neck squamous cell carcinoma and refinement of 3q amplifica-tion. Cancer Res. 61(11), 4506–4513 (2001)

18. C.A. Marcos, M. Alonso-Guervos, N.R. Prado, T.S. Gimeno, F.D.Iglesias, M. Hermsen, J.L. Llorente, Genetic model of transformationand neoplastic progression in laryngeal epithelium. Head Neck33(2), 216–224 (2010)

19. N. Uzawa, D. Akanuma, A. Negishi, H. Iwaki, Y. Uzawa, T.Amagasa, M.A. Yoshida, Homozygous deletions on the short armof chromosome 3 in human oral squamous cell carcinomas. OralOncol. 37(4), 351–356 (2001)

20. M.A. Pershouse, A.K. El-Naggar, K. Hurr, H. Lin, W.K. Yung, P.A.Steck, Deletion mapping of chromosome 4 in head and neck squa-mous cell carcinoma. Oncogene 14(3), 369–373 (1997)

21. X.L. Wang, K. Uzawa, F.L. Imai, H. Tanzawa, Localization of anovel tumor suppressor gene associated with human oral cancer onchromosome 4q25. Oncogene 18(3), 823–825 (1999)

22. E.J. Mao, S.M. Schwartz, J.R. Daling, A.M. Beckmann, Loss ofheterozygosity at 5q21-22 (adenomatous polyposis coli gene region)in oral squamous cell carcinoma is common and correlated withadvanced disease. J. Oral Pathol. Med. 27(7), 297–302 (1998)

23. R.A. Brekken, J.P. Overholser, V.A. Stastny, J. Waltenberger, J.D.Minna, P.E. Thorpe, Selective inhibition of vascular endothelialgrowth factor (VEGF) receptor 2 (KDR/Flk-1) activity by a mono-clonal anti-VEGF antibody blocks tumor growth in mice. CancerRes. 60(18), 5117–5124 (2000)

24. L.C. da Silva Veiga, N.A. Bergamo, P.P. dos Reis, L.P. Kowalski,S.R. Rogatto, DNA gains at 8q23.2: a potential early marker in headand neck carcinomas. Cancer Genet. Cytogenet. 146(2), 110–115(2003)

25. S.C. Lin, Y.J. Chen, S.Y. Kao, M.T. Hsu, C.H. Lin, S.C. Yang, T.Y.Liu, K.W. Chang, Chromosomal changes in betel-associated oralsquamous cell carcinomas and their relationship to clinical parame-ters. Oral Oncol. 38(3), 266–273 (2002)

26. C. Garnis, B.P. Coe, A. Ishkanian, L. Zhang, M.P. Rosin, W.L. Lam,Novel regions of amplification on 8q distinct from the MYC locusand frequently altered in oral dysplasia and cancer. GeneChromosome Cancer 39(1), 93–98 (2004)

27. C. Garnis, B.P. Coe, L. Zhang, M.P. Rosin, W.L. Lam, Overexpressionof LRP12, a gene contained within an 8q22 amplicon identified byhigh-resolution array CGH analysis of oral squamous cell carcinomas.Oncogene 23(14), 2582–2586 (2004)

28. J.D. Cha, H.J. Kim, I.H. Cha, Genetic alterations in oral squamouscell carcinoma progression detected by combining array-based com-parative genomic hybridization and multiplex ligation-dependentprobe amplification. Oral Surg. Oral Med. Oral Pathol. Oral Radiol.Endod. 111(5), 594–607 (2011)

29. S.M. Gollin, Chromosomal alterations in squamous cell carcinomasof the head and neck: window to the biology of disease. Head Neck23(3), 238–253 (2001)

30. A. Sparano, K.M. Quesnelle, M.S. Kumar, Y. Wang, A.J. Sylvester,M. Feldman, D.A. Sewell, G.S.Weinstein, M.S. Brose, Genome-wideprofiling of oral squamous cell carcinoma by array-based comparativegenomic hybridization. Laryngoscope 116(5), 735–741 (2006)

38 I.P. Ribeiro et al.

31. S.L. Cooke, J.C. Pole, S.F. Chin, I.O. Ellis, C. Caldas, P.A. Edwards,High-resolution array CGH clarifies events occurring on 8p in carci-nogenesis. BMC Cancer 8 , 288 (2008)

32. M.J. Worsham, G. Pals, J.P. Schouten, R.M. Van Spaendonk, A.Concus, T.E. Carey, M.S. Benninger, Delineating genetic pathwaysof disease progression in head and neck squamous cell carcinoma.Arch. Otolaryngol. Head Neck Surg. 129(7), 702–708 (2003)

33. J.W. Rocco, D. Sidransky, p16(MTS-1/CDKN2/INK4a) in cancerprogression. Exp. Cell Res. 264(1), 42–55 (2001)

34. P. van der Riet, H. Nawroz, R.H. Hruban, R. Corio, K. Tokino, W.Koch, D. Sidransky, Frequent loss of chromosome 9p21-22 early inhead and neck cancer progression. Cancer Res. 54(5), 1156–1158(1994)

35. C. Jin, Y. Jin, D. Gisselsson, J. Wennerberg, T.S.Wah, B. Stromback,Y.L. Kwong, F. Mertens, Molecular cytogenetic characterization ofthe 11q13 amplicon in head and neck squamous cell carcinoma.Cytogenet Genome Res 115(2), 99–106 (2006)

36. R.A. Parikh, J.S. White, X. Huang, D.W. Schoppy, B.E. Baysal, R.Baskaran, C.J. Bakkenist, W.S. Saunders, L.C. Hsu, M. Romkes,S.M. Gollin, Loss of distal 11q is associated with DNA repairdeficiency and reduced sensitivity to ionizing radiation in head andneck squamous cell carcinoma. Gene Chromosome Cancer 46(8),761–775 (2007)

37. K. Freier, C. Sticht, C. Hofele, C. Flechtenmacher, D. Stange, L.Puccio, G. Toedt, B. Radlwimmer, P. Lichter, S. Joos, Recurrentcoamplification of cytoskeleton-associated genes EMS1 andSHANK2 with CCND1 in oral squamous cell carcinoma. GeneChromosome Cancer 45(2), 118–125 (2006)

38. H. Mineta, K. Miura, S. Takebayashi, Y. Ueda, K. Misawa, H.Harada, J. Wennerberg, M. Dictor, Cyclin D1 overexpression corre-lates with poor prognosis in patients with tongue squamous cellcarcinoma. Oral Oncol. 36(2), 194–198 (2000)

39. A. Namazie, S. Alavi, O.I. Olopade, G. Pauletti, N. Aghamohammadi,M. Aghamohammadi, J.A. Gornbein, T.C. Calcaterra, D.J. Slamon,M.B. Wang, E.S. Srivatsan, Cyclin D1 amplification and p16(MTS1/CDK4I) deletion correlate with poor prognosis in head and necktumors. Laryngoscope 112(3), 472–481 (2002)

40. Z. Yu, P.M. Weinberger, B.G. Haffty, C. Sasaki, C. Zerillo, J. Joe, D.Kowalski, J. Dziura, R.L. Camp, D.L. Rimm, A. Psyrri, Cyclin D1 isa valuable prognostic marker in oropharyngeal squamous cell carci-noma. Clin. Cancer Res. 11(3), 1160–1166 (2005)

41. U. Duvvuri, D.J. Shiwarski, D. Xiao, C. Bertrand, X. Huang, R.S.Edinger, J.R. Rock, B.D. Harfe, B.J. Henson, K. Kunzelmann, R.Schreiber, R.S. Seethala, A.M. Egloff, X. Chen, V.W. Lui, J.R.Grandis, S.M. Gollin, TMEM16A induces MAPK and contributes

directly to tumorigenesis and cancer progression. Cancer Res.72(13), 3270–3281 (2012)

42. O. Nilsson, Profiling of ileal carcinoids. Neuroendocrinology 97(1),7–18 (2013)

43. M.H. Mossink, A. van Zon, R.J. Scheper, P. Sonneveld, E.A.Wiemer, Vaults: a ribonucleoprotein particle involved in drug resis-tance? Oncogene 22(47), 7458–7467 (2003)

44. N. Pecina-Slaus, M. Kljaic, T. Nikuseva-Martic, Loss of heterozy-gosity of APC and CDH1 genes in laryngeal squamous cell carcino-ma. Pathol. Res. Pract. 201(8–9), 557–563 (2005)

45. D. Hardisson, C. Alvarez-Marcos, A. Salas-Bustamante, M. Alonso-Guervos, N. Sastre, A. Sampedro, Numerical aberrations of chromo-somes 8, 9, 11, and 17 in squamous cell carcinoma of the pharynx andlarynx: a fluorescence in situ hybridization and DNA flow cytometricanalysis of 50 cases. Oral Oncol. 40(4), 409–417 (2004)

46. C. Scully, J.K. Field, H. Tanzawa, Genetic aberrations in oral or headand neck squamous cell carcinoma (SCCHN): 1. Carcinogen metab-olism, DNA repair and cell cycle control. Oral Oncol. 36(3), 256–263 (2000)

47. A. Carneiro, A. Isinger, A. Karlsson, J. Johansson, G. Jonsson, P.O.Bendahl, D. Falkenback, B. Halvarsson, M. Nilbert, Prognostic im-pact of array-based genomic profiles in esophageal squamous cellcancer. BMC Cancer 8 , 98 (2008)

48. C.L. Martin, S.C. Reshmi, T. Ried, W. Gottberg, J.W. Wilson, J.K.Reddy, P. Khanna, J.T. Johnson, E.N. Myers, S.M. Gollin,Chromosomal imbalances in oral squamous cell carcinoma: exami-nation of 31 cell lines and review of the literature. Oral Oncol. 44(4),369–382 (2008)

49. B.J.M. Braakhuis, E. Bloemena, C.R. Leemans, R.H. Brakenhoff,Molecular analysis of surgical margins in head and neck cancer: morethan a marginal issue. Oral Oncol. 46(7), 485–491 (2010)

50. J. Califano, P. van der Riet, W. Westra, H. Nawroz, G. Clayman, S.Piantadosi, R. Corio, D. Lee, B. Greenberg, W. Koch, D. Sidransky,Genetic progression model for head and neck cancer: implications forfield cancerization. Cancer Res. 56(11), 2488–2492 (1996)

51. A. Bahl, P. Kumar, L. Dar, B.K. Mohanti, A. Sharma, A. Thakar, V.Karthikeyan, K. Sikka, C. Singh, K. Poo, J. Lodha, Prevalence andtrends of human papillomavirus in oropharyngeal cancer in a pre-dominantly north Indian population. Head Neck, (2013)

52. M.L. Gillison, Human papillomavirus-associated head and neck can-cer is a distinct epidemiologic, clinical, and molecular entity. Semin.Oncol. 31(6), 744–754 (2004)

53. C.R. Leemans, B.J. Braakhuis, R.H. Brakenhoff, The molecularbiology of head and neck cancer. Nat. Rev. Cancer 11(1), 9–22(2011)

Genetic gains and losses in oral squamous cell carcinoma 39