Field cancerization is a potential mechanism for SCCHN tumorigenesis. Many patients diagnosed with SCCHN have had chronic exposures to tobacco and alcohol and are at increased risk of local recurrence and second primaries. In 1953, Slaughter et al. described the mucosa of the upper aerodigestive tract of such patients to be “condemned” since large areas are chronically exposed to the environmental mutagens and thus, are all at risk for developing dysplasias, and subsequently, carcinomas [6
]. Such phenomenon can provide a pathogenetic basis for the occurrence of synchronous and metachronous primaries.
In 1990, Vogelstein et al. was the first to propose a multistage pathogenesis for cancer, specifically of the colon. This genetic model for colorectal tumorigenesis included the activation of oncogenes and inactivation of tumor suppressor genes. He described that multiple “hits” (at least four mutations) were necessary for malignant transformation. Step-wise advances have been suggested including initiation, promotion and progression. This schema has been observed in many other cancers including those of the brain, bladder, in addition to head and neck. Califano et al. reported a “Vogelgram” for SCCHN, which linked histologic features in the progression of SCCHN to specific molecular alterations [7
]. They proposed that it is the accumulation of genetic events, not the order, that determines progression.
Further investigation focused on the transcriptional changes that occur in the progression from normal-appearing mucosa to dysplastic tissue and invasive SCCHN [8
]. Microarray analysis of RNA isolated from a continuum of such specimens identified genes with differential expression patterns. Similarly to the genetic progression model of SCCHN, this group showed that most of the alterations in their transcriptional progression model occurred before the development malignancy. Identification of genes that were significantly upregulated included integrin α
2.1. Cytogenetic alterations
Genomic changes that occur in premalignant and malignant lesions can manifest themselves on different levels (chromosomal, DNA, RNA, and protein) and in a variety of ways such as point mutations, amplifications, deletions, and chromosomal alterations. Common methods to screen for the presence of these changes include, but are not limited to, conventional cytogenetics, comparative genomic hybridization (CGH), spectral karyotyping, and cDNA microarrays. CGH is a cytogenetic technique that assesses for gains or losses in DNA content (e.g., chromosomal imbalances) within a tumor’s entire genome. Tumor DNA can be obtained by microdissection of an H&E-stained specimen, while DNA from peripheral blood leukocytes serves as a control. CGH is performed by hybridizing the tumor DNA and normal DNA, which are labeled with different fluorescent tags, to normal human metaphase preparations. Epifluorescent microscopy and image analysis is employed to detect DNA gains or losses as compared to controls. CGH does not show structural changes between chromosomes. The minimal size of chromosome that can be used is similar to conventional cytogenetics (3–5 Mb).
Weber et al. employed this technique to determine the average number of chromosomal imbalances in 12 oral premalignant lesions (including dysplasias and carcinomas-in situ) and 14 invasive oral SCCHNs [9
]. An average of 3.2 ± 1.2 imbalances were seen in premalignant lesions while invasive SCCHNs had a significantly higher average of 11.9 ± 1.9 (p
= 0.003) imbalances. Within the premalignant lesions, gains were identified on 8q and 16p while losses were found on 3p, 5q, 13q, and 4q, 8p, and 9p. In individual biopsies from the same subject that contained both premalignancy and invasive carcinoma, most of the genomic alterations found in premalignancy were also found in SCCHN.
Brieger et al. used CGH to analyze chromosomal alterations in 22 oropharyngeal SCCHN tumors and their surrounding benign mucosa [10
]. In the morphologically benign mucosa collected 2 cm from the primary tumor, no chromosomal alterations were identified. In benign mucosa located 1 cm from the tumor, the most common amplifications were in 15q and 21q. Nearly all of these changes were seen in the primary tumor as well. The authors speculated that some second primaries may be explained by a monoclonal model followed by lateral epithelial spread. Kim et al. utilized chromosome in-situ hybridization (CISH) targeting chromosomes 9 and 17 to characterize oral lichen planus lesions [11
]. When mucosal epithelium was compared to control lymphocytes, a higher proportion of polysomic and monosomic cells for chromosome 9 was found (p
These cytogenetic findings are consistent with previous reports implicating molecular changes early on in head and neck tumorigenesis before histologic and phenotypic changes, in addition to their accumulation through successive stages. CGH continues to be an invaluable tool, but the other aforementioned techniques are also widely used.
2.2. Molecular alterations
2.2.1. Microsatellite instability
Microsatellites are repeats of non-coding DNA sequences that occur normally within the human genome. Defects in the DNA repair process can lead to microsatellites that are abnormally short or long; this process has been termed microsatellite instability (MI). MI is indirect evidence of an abnormal mismatch repair (MMR) protein’s function (hMLH1, PMS2, MSH2, MSH6). A proposed mechanism relevant in SCCHN tumorigenesis is through promoter hypermethylation. When MMR promoters are hypermethylated, it provides indirect evidence of a higher chance that promoters of tumor suppressor genes are hypermethylated too, and therefore nonfunctional [12
]. Alternatively, when a microsatellite repeat replication error goes uncorrected, a germ line hereditary mutation could result leading to inactivation of tumor suppressor genes and uncontrolled cell and tumor growth. This concept of a mutator phenotype provides an alternative to a multistage accumulation of genetic alterations to explain head and neck tumorigenesis. Specifically, the loss of function of a gene critical for the repair of DNA damage greatly increases the mutation rate at other loci leading to genome-wide instability.
In a study of 93 premalignant and 18 invasive SCCHN cases, an increasing trend of MI was found from hyperplasias (6% of specimens) to dysplasias/CIS (27%) and to invasive cancers (33%) [13
]. A similar trend was found in another study where 15% of dysplasias and 30% of invasive cases manifested MI at multiple loci [14
]. Partridge et al. found as high as 55% of 31 leukoplakias and erythroplakias to show MI [15
]. The incidences of MI found in these and other studies in head and neck malignancies are significantly higher than those reported in breast, skin and non-small-cell lung cancers. The prevalence of MI appears to vary between tumor types.
2.2.2. Loss of heterozygosity (LOH)
An allele of a gene, such as a tumor suppressor gene, can be inactivated by mutation. When this occurs in a parent’s germline cell, the inactivated allele is passed onto its offspring resulting in heterozygosity. When genomic loss occurs in the somatic cell of the offspring affecting the remaining allele, LOH occurs and tumor suppressive function in that cell is lost. LOH assays commonly employ microsatellite analysis to assess polymorphic chromosomal regions that map in or near tumor suppressor genes.
It is known that LOH at 9p21 and/or 3p14 in premalignant oral lesions increases their probability of malignant transformation [16
]. Weber et al. identified LOH at 3p, 9p, 17p, and 18q in dysplasias and 11q, 13q, and 14q in CIS [9
]. Other chromosomal losses have been associated with increased risk: 4q, 8p, 11q, 13q, and 17p. Hyperplastic or dysplastic lesions with LOH at 3p and/or 9p plus one of the other above losses were found to have a 33-fold increase in cancer risk [17
]. In lesions that lack significant dysplasia to guide estimation of cancer risks, molecular markers such as these may yet prove to be useful. Zhang et al. proposed a staging system incorporating assessments from both histology and LOH criteria [18
p53 is a well-characterized tumor suppressor gene located on chromosome 17p. Its function is to regulate cell growth arrest and apoptosis. The p53 protein has a very short halflife, and thus, is difficult to detect in benign tissues. Overexpression of p53 can result from mutation, a defect in its degradation, or from binding to other proteins. On the whole, mutations of the p53 tumor suppressor gene have been found in about half of SCCHN tumors and are the most common genetic alteration found in human malignancy. The types of mutations vary including point mutations, transversions, transitions, and deletions. The most common type are missense mutants, which have a longer halflife than the wild-type form. Various proteins are known to bind with p53 such as SV40 large T, which blocks its DNA binding capability. Binding with adenovirus E1B blocks p53’s transcriptional activity. Finally, binding with HPV E6 targets p53 for accelerated degradation.
During progressive steps in SCCHN carcinogenesis, increasing frequencies of p53 alterations and genomic instability have been identified. Shin et al. analyzed p53 expression and chromosomal polysomy via immunohistochemistry and chromosome insitu hybridization, respectively, in 136 epithelial specimens. 19% of adjacent normal-appearing mucosa, 29% of hyper-plastic lesions, 46% of dysplastic lesions, and 58% of SCCHN tumors expressed p53. Normal-appearing mucosa lacked detectable p53 levels as expected. p53-positive dysplastic lesions showed a 2 to 4-fold increase in the level of chromosome 9 and 17 polysomy (p
= 0.002 and 0.005, respectively). In the remaining groups, increasing trends were found with histological progression [19
]. These findings suggest that premalignancy is associated with altered p53 expression and increased genomic instability, which may be early markers of carcinogenesis.
A Dutch study attempted to correlate p53 staining patterns with risk of malignant transformation. Thirty-five premalignant lesions were followed for a range of 1–16 years. Staining patterns included suprabasal or basal expression or no expression at all. The positive predictive value of suprabasal p53 staining progressing into carcinoma was 86% whereas the negative predictive value of basal or absent p53 expression was 82% [20
The implications of p53-positivity in histologically normal-appearing mucosa are incompletely understood. Some attribute it as an effect of field cancerization; however, it is unknown whether any prognostic significance is associated with this finding. Ogden et al. examined specimens from 13 patients with p53-positive normal-appearing mucosa and compared their rate of second primaries to those of 9 patients with p53-negative normal-appearing mucosa [21
]. No significant difference was seen between the groups after a mean follow-up of 4.75 years and 4.1 years, respectively.
2.2.4. Retinoic acid receptor-β
Retinoids (natural and synthetic derivatives of vitamin A) regulate cell growth and differentiation and have growth-suppressive effects in epithelial cells. Retinoic acid receptor-β (RAR-β) is a steroid hormone receptor whose expression is suppressed in premalignant and malignant head and neck lesions via an unknown mechanism. Restoration of RAR-β expression, e.g., with isoretinoin or 13-cis retinoic acid, seems to correlate with the growth-inhibitory effect of retinoids in solid tumors and premalignancies. However, as will be discussed later, isoretinoin failed to show efficacy as monotherapy in chemoprevention and is being studied in combination with other biologic agents (see Section 5.2. Biochemoprevention).
2.2.5. Reactivation of telomerase
Telomeres are DNA repeats that cap the ends of chromosomes for stability purposes. The enzyme telomerase replenishes telomeres as they necessarily get consumed by each replicative cycle. Telomerase has been detected in germline cells, which allows them to divide repeatedly. However, most postnatal somatic cells express it only at very low levels in a cell-cycle dependent manner. Increased telomerase activity is usually associated with longer cell survival. Also, since it is activated in many human cancers and is undetectable in most normal tissues, telomerase makes for an attractive potential tumor marker.
Patel et al. utilized a TRAP (telomerase rapid amplification protocol) assay to identify telomerase activation in 78% of 110 SCCHN tumors and 85% of 36 precancerous lesions [22
]. 53% of adjacent normal-appearing tissues (1 cm away from tumor) were also positive. Additionally, mean telomere length was shorter in malignant tissues as compared to adjacent normal-appearing tissues. Patients with telomerase activation in adjacent normal-appearing tissues and patients with higher telomere length in malignant tissues were noted to have poor disease-free survival. Similar to LOH, telomerase activation appears to be an early event in carcinogenesis.
Liao et al. used a modified TRAP assay to identify telomerase activity in 39% of normal-appearing mucosal samples, 55% of leukoplakias, and 82% of oral SCCHN samples [23
]. Zhang et al. developed a reproducible liquid scintillation counter to detect low radioactive counts in all normal-appearing samples and high counts in 86% of malignant head and neck tumors. In ten dysplastic specimens, two with severe dysplasia showed higher telomerase activities than those with mild dysplasia [24
]. Sumida et al. also used a modified TRAP assay to detect telomerase activity in 71% of dysplastic lesions. RTPCR was employed to measure the expression levels of the three telomerase components–hTR, TP1, and hTRT. Detection of hTRT-mRNA by RT-PCR appeared to be more sensitive for telomerase than measurement by the TRAP assay [25
2.2.6. p16/cyclin D1/Rb
In addition to p53, Rb is the other major tumor suppression pathway in human carcinogenesis. Rb normally suppresses cell growth by binding to E2F1 transcription factors and preventing cells from advancing from G1 into S phase of the cell cycle (G1 arrest). The proto-oncogenes, cyclin D-dependent kinases CDK4 and CDK6, phosphorylate and inactivate Rb thereby allowing for cell cycle progression. Cyclin D1 facilitates CDK complex formation and pRb binding. p16IN K4a
is one type of cyclin-dependent kinase (CDK) inhibitor, which binds to CDK4 and CDK6, thereby activating Rb and G1 arrest. The major control point of p16 is in the methylation of its promoter. Another CDK inhibitor class includes p21waf-1
, whose role also includes inhibiting DNA replication and inducing cell cycle arrest. Altered expression of pRb and p16 has been reported in oral carcinomas [26
]. In fact, alterations in p16 are considered the most common genetic alteration in SCCHN [27
]. Cyclin D1 amplification and overexpression have been also identified [26
]. The HPV oncoprotein E7 is known to inactivate Rb.
Soni et al. performed IHC analysis of p16, cyclin D1, and Rb expression in 81 normal-appearing oral mucosal specimens, 52 hyperplastic and 38 dysplastic lesions and 220 SCCHN tumors. 11% of hyperplastic and dysplastic and 18% of SCCHN tumors revealed alterations in all three genes. Loss of p16 was identified as the earliest event. Loss of Rb was associated with transition from hyperplasia to dysplasia. Rb loss with cyclin D1 overexpression or p53 overexpression alone was associated with transition to SCCHN. Loss of Rb along with p53 overexpression was a significant predictor of survival.
2.2.7. Mitochondrial DNA
As opposed to alterations in human genomic DNA, changes can occur in mitochondrial DNA instead. A common location for polymorphisms and mutations is the poly-cytosine tract (C-tract) of the displacement loop of the mitochondrial genome. Ha et al., tested 137 premalignant lesions for C-tract DNA alterations using PCR and polyacrylamide gel electrophoresis [28
]. When compared to controls, 37% of patients exhibited C-tract alterations. Moreover, the occurrence of C-tract changes increased with increasing severity of dysplasia up to carcinoma in-situ. An identical pattern was seen in 8 out of 10 patients with multiple, synchronous tumors as well as in 2 of 3 patients with metachronous tumors. These findings suggest that C-tract alterations may be markers of clonality when multiple tumors occur within the same patient.
The epidermal growth factor receptor (EGFR) pathway has been investigated due its role in SCCHN tumorigenesis. Epidermal growth factor receptor (EGFR) and its primary autocrine ligand, transforming growth factor-α
), are upregulated early during carcinogenesis via gene amplification and/or transcriptional activation. Overexpression of EGFR is present in malignant, premalignant and normal-appearing tissues from SCCHN patients and is correlated with poor prognosis [29
]. Staining of TGF-α
in the statum germinativum increased linearly in oral leukoplakia with mild, moderate, and severe dysplasia as compared to control mucosa. EGFR staining increased linearly in the stratum spinosum in oral leukoplakia with increasing degrees of dysplasia [31
Upon binding of ligand, EGFR dimerizes and its intracellular tyrosine kinase (e.g., JAK2) activates and phosphorylates several downstream molecules, one of which is the STAT3 transcription factor. STAT3 contributes to a diverse array of cellular responses, including proliferation, differentiation, and suppression of apoptosis. Phosphorylated STAT3 dimerizes and translocates from the cytoplasm to the nucleus. In the nucleus, the phosphorylated/activated STAT3 dimer binds to STAT3 DNA response elements in the promoters of target genes, stimulating their transcription and gene expression. It is the repertoire of STAT3 target genes that are induced in a particular cell type helps to determine the cellular response to growth factor and cytokine stimulation. Overexpression of EGFR is also correlated with constitutive activation of STAT3 via an antiapoptotic mechanism (e.g., Bcl-xL
). Inhibition of STAT3 via anti-sense strategies increases apoptosis and decreases Bcl-xL
expression in a xenograft model [30
]. Early activation of EGFR and STAT3 in carcinogenesis implicates this pathway as potential therapeutic target in premalignancy.