Targeting the genetic landscape of oral potentially malignant disorders has the potential as a preventative strategy in oral cancer.

This study reviews the molecular landscape of oral potentially malignant disorders (OPMD). We examine the impact of tumour heterogeneity, the spectrum of driver mutations (TP53, CDKN2A, TERT, NOTCH1, AJUBA, PIK3CA, CASP8) and gene transcription on tumour progression. We comment on how some of these mutations impact cellular senescence, field cancerization and cancer stem cells. We propose that OPMD can be monitored more closely and more dynamically through the use of liquid biopsies using an appropriate biomarker of transformation. We describe new gene interactions through the use of a systems biology approach and we highlight some of the first studies to identify functional genes using CRISPR-Cas9 technology. We believe that this information has translational implications for the use of re-purposed existing drugs and/or new drug development. Further, we argue that the use of digital technology encompassing clinical and laboratory-based data will create relevant datasets for machine learning/artificial intelligence. We believe that therapeutic intervention at an early molecular premalignant stage should be an important preventative strategy to inhibit the development of oral squamous cell carcinoma and that this approach is applicable to other aerodigestive tract cancers.

Cancer-associated fibroblasts regulate keratinocyte cell-cell adhesion via TGF-ß-dependent pathways in genotype-specific oral cancer.

The interrelationship between malignant epithelium and the underlying stroma is of fundamental importance in tumour development and progression. In the present study, we used cancer-associated fibroblasts (CAFs) derived from genetically unstable oral squamous cell carcinomas (GU-OSCC), tumours that are characterized by the loss of genes such as TP53 and p16INK4A and with extensive loss of heterozygosity, together with CAFs from their more genetically stable (GS) counterparts that have wild-type TP53 and p16INK4A and minimal loss of heterozygosity (GS-OSCC). Using a systems biology approach to interpret the genome-wide transcriptional profile of the CAFs, we show that transforming growth factor-ß (TGF-ß) family members not only had biological relevance in silico but also distinguished GU-OSCC-derived CAFs from GS-OSCC CAFs and fibroblasts from normal oral mucosa. In view of the close association between TGF-ß family members, we examined the expression of TGF-ß1 and TGF-ß2 in the different fibroblast subtypes and showed increased levels of active TGF-ß1 and TGF-ß2 in CAFs from GU-OSCC. CAFs from GU-OSCC, but not GS-OSCC or normal fibroblasts, induced epithelial-mesenchymal transition and down-regulated a broad spectrum of cell adhesion molecules resulting in epithelial dis-cohesion and invasion of target keratinocytes in vitro in a TGF-ß-dependent manner. The results demonstrate that the TGF-ß family of cytokines secreted by CAFs derived from genotype-specific oral cancer (GU-OSCC) promote, at least in part, the malignant phenotype by weakening intercellular epithelial adhesion.

Fibroblast activation and senescence in oral cancer.

There is now compelling evidence that the tumour stroma plays an important role in the pathogenesis of cancers of epithelial origin. The pre-eminent cell type of the stroma is carcinoma-associated fibroblasts. These cells demonstrate remarkable heterogeneity with activation and senescence being common stress responses. In this review, we summarise the part that these cells play in cancer, particularly oral cancer, and present evidence to show that activation and senescence reflect a unified programme of fibroblast differentiation. We report advances concerning the senescent fibroblast metabolome, mechanisms of gene regulation in these cells and ways in which epithelial cell adhesion is dysregulated by the fibroblast secretome. We suggest that the identification of fibroblast stress responses may be a valuable diagnostic tool in the determination of tumour behaviour and patient outcome. Further, the fact that stromal fibroblasts are a genetically stable diploid cell population suggests that they may be ideal therapeutic targets and early work in this context is encouraging.

Senescent cancer-associated fibroblasts secrete active MMP-2 that promotes keratinocyte dis-cohesion and invasion.

BACKGROUND: Previous studies have demonstrated that senescent cancer-associated fibroblasts (CAFs) derived from genetically unstable oral squamous cell carcinomas (GU-OSCC), unlike non-senescent CAFs from genetically stable carcinomas (GS-OSCC), promoted keratinocyte invasion in vitro in a paracrine manner. The mechanism by which this occurs is unclear. METHODS: Previous work to characterise the senescent-associated secretory phenotype (SASP) has used antibody arrays, technology that is limited by the availability of suitable antibodies. To extend this work in an unbiased manner, we used 2D gel electrophoresis and mass spectroscopy for protein identification. Matrix metalloproteinases (MMPs) were investigated by gelatin zymography and western blotting. Neutralising antibodies were used to block key molecules in the functional assays of keratinocyte adhesion and invasion. RESULTS: Among a variety of proteins that were differentially expressed between CAFs from GU-OSCC and GS-OSCC, MMP-2 was a major constituent of senescent CAF-CM derived from GU-OSCC. The presence of active MMP-2 was confirmed by gelatine zymography. MMP-2 derived from senescent CAF-CM induced keratinocyte dis-cohesion and epithelial invasion into collagen gels in a TGF-ß-dependent manner. CONCLUSIONS: Senescent CAFs from GU-OSCC promote a more aggressive oral cancer phenotype by production of active MMP-2, disruption of epithelial adhesion and induction of keratinocyte invasion.

Increased secretion of tissue inhibitors of metalloproteinases 1 and 2 (TIMPs -1 and -2) in fibroblasts are early indicators of oral sub-mucous fibrosis and ageing.

Oral submucous fibrosis (OSMF) is associated with paan chewing, altered collagen metabolism, inflammation and the upregulation of numerous cytokines. OSMF fibroblasts accumulate senescent cells at an increased rate because of increased reactive oxygen species production and DNA double-strand breaks (DDBs), generated intrinsically by damaged mitochondria. This results in a reduced replicative lifespan. However, it is still unclear which other changes are intrinsic to the fibroblasts and associated with OSMF rather than the paan chewing habit or the OSMF environment. Both the oral epithelium and the mesenchyme have elevated levels of TGF-ß(1) in OSMF in vivo. However, in cultured fibroblasts, secreted levels of TGF-ß(1,) other cytokines and the matrix metalloproteinases 1 and 2 showed no association with OSMF. In contrast, the tissue inhibitors of metalloproteinases, TIMP-1 and TIMP-2, were increased in 10/11 OSMF fibroblast cultures relative to normal and non-diseased paan user controls. OSMF fibroblast collagen levels were normal. TIMP levels correlated with replicative lifespan of the cultures but not with the presence of senescent cells, as senescent cell depletion in OSMF fibroblast cultures did not result in a reduction in either TIMP-1 or TIMP-2. However, the introduction of unrepairable DDBs into normal oral fibroblasts by ionizing radiation increased TIMP-1 and TIMP-2 secretion by two-fold and seven-fold, respectively, within 5 days, replicating early senescence and the elevation seen in OSMF cultures. Therefore, increased fibroblast TIMP-1/2 levels could be early disease-specific markers of OSMF onset, DDBs and ageing and may have clinical significance, as OSMF can be reversed in its early stages.

Association of increased telomere lengths in limited scleroderma, with a lack of age-related telomere erosion.

OBJECTIVES: Telomere erosion, a feature of biological ageing, is implicated in a wide range of diseases. Its impact on autoimmune diseases remains unclear although autoantibodies against many telomere nucleoprotein components are prevalent in these diseases. We aimed to assess if telomere biology was abnormal in a cohort of patients with limited cutaneous systemic sclerosis (lcSSc). METHODS: Telomere lengths in peripheral blood leucocytes (PBL) were determined using Southern blotting methods in a cohort of lcSSc subjects (n=43; age range 37-80 years) and a control population (n=107; age range 21-65 years). RESULTS: Telomere lengths in lcSSc subjects were longer than controls (p<0.001), did not show age-related telomere erosion and differed significantly from age-matched controls only after 50 years of age (p<0.001). CONCLUSIONS: This is the first report of maintenance of telomere lengths in an autoimmune disease state. These data indicate aberrant telomere biology and irregular biological ageing from the fifth decade of life. These findings provide insight into compromised DNA damage repair in lcSSc. Whether these observations indicate a causal or consequential relationship requires further investigation. This in turn, may provide potential novel targets for therapeutic intervention.

Upregulation of HIF-1alpha in malignant transformation of oral submucous fibrosis.

BACKGROUND: Oral submucous fibrosis (OSF) is a precancerous condition showing extensive fibrosis of the submucosa and affects most parts of the oral cavity, including pharynx and upper third of the oesophagus. The molecules involved in the biological pathways of the fibrotic process appeared to be either down- or upregulated at different stages of the disease. Despite the precancerous nature, malignant transformation of the epithelium in the background of fibrosis has not been studied in detail. HIF-1alpha is a known transcription factor that is induced by hypoxia. AIMS: To test the hypothesis that hypoxia plays a role in malignant transformation and progression of OSF. MATERIALS AND METHODS: We used both formalin-fixed and frozen samples of OSF and normal mucosa to investigate the relationship between HIF-1alpha and epithelial dysplasia using immunohistochemistry and RT-PCR. CONCLUSIONS: Our data indicate that HIF-1alpha is upregulated at both protein and mRNA levels in OSF and the correlation with epithelial dysplasia is statistically significant (P < 0.001). We propose that HIF-1alpha may play a role in malignant transformation of OSF. Further, over-expression of HIF-1alpha may contribute to the progression of fibrosis. It may be possible to use HIF-1alpha as a marker for malignant transformation of OSF.

Dissecting the non-canonical functions of telomerase.

It is now well established that the canonical function of telomerase protects the telomere repeats from erosion and the consequent induction of replicative senescence or apoptosis. In the absence of key cell cycle checkpoint proteins, the canonical function of telomerase also prevents chromosome fusions and immortalizes human cells. The canonical function of telomerase requires both the telomerase reverse transcriptase enzyme (TERT) which adds telomere (TTAGGG) repeats to the chromosome ends and the telomerase RNA component (TERC), which provides the template for TERT. However, there is growing evidence that telomerase has other (non-canonical) functions. These functions can be divided further into those that require telomerase activity but not telomere lengthening (non-canonical I or NC I) and those that require neither telomerase activity nor telomere lengthening (non-canonical II or NC II). NC I functions are associated with the induction of neoplasia in both epidermis and mammary gland, the correct response to DNA damage, and insensitivity to transforming growth factor beta. In contrast, NC II functions are not sufficient for the induction of neoplasia and are associated with the activation of the WNT and MYC signaling pathways in keratinocytes and a more general resistance to the induction of apoptosis by a variety of stimuli. The overexpression of either TERT or TERC appears to be capable of providing NC I functions but NC II functions require neither TERC nor the integrity of the TERT catalytic site. The molecular mechanisms underpinning both NC I and NC II are largely obscure but transcriptional profile changes have been reported by some groups. In this article, we will discuss the proposed mechanisms of NC I and NC II and their relevance to normal and neoplastic cell functions.

Human squamous cell carcinomas lose a mortality gene from chromosome 6q14.3 to q15.

Normal human keratinocytes possess a finite replicative lifespan. Most advanced squamous cell carcinomas (SCCs), however, are immortal, a phenotype that is associated with p53 and INK4A dysfunction, high levels of telomerase and loss of heterozygosity (LOH) at several genetic loci, suggestive of the dysfunction of other mortality genes. We show here that human chromosome 6 specifically reduces the proliferation or viability of a human SCC line, BICR31, possessing LOH across the chromosome. This was determined by an 88% reduction in colony yield (P<0.001), following the reintroduction of an intact normal chromosome 6 by monochromosome transfer. Deletion analysis of immortal segregants using polymorphic markers revealed the loss of a 2.9 Mbp interval, centred on marker D6S1045 at 6q14.3-q15, in 6/19 segregants. Crucially, allelic losses of this region were not identified in control hybrids constructed between chromosome 6 and the BICR6 SCC cell line that is heterozygous for chromosome 6 and which showed no reduction in colony formation relative to the control chromosome transfers. This indicates that the minimally deleted region at D6S1045 is not the result of fragile sites, a recombination hot spot, or a feature of the monochromosome transfer technique. LOH of D6S1045 was found in 2/9 immortal SCC lines and was part of a minimally deleted region of line BICR19. Furthermore, allelic imbalance, consistent with LOH, was detected in 3/17 advanced SCCs of the tongue. These results suggest the existence of a suppressor of SCC immortality and tumour development at chromosome 6q14.3-q15, which is important to a subset of human SCCs.

Human fibroblast replicative senescence can occur in the absence of extensive cell division and short telomeres.

Ectopic expression of telomerase blocks both telomeric attrition and senescence, suggesting that telomeric attrition is a mitotic counting mechanism that culminates in replicative senescence. By holding human fibroblast cultures confluent for up to 12 weeks at a time, we confirmed previous observations and showed that telomeric attrition requires cell division and also, that senescence occurs at a constant average telomere length, not at a constant time point. However, on resuming cell division, these long-term confluent (LTC) cultures completed 15-25 fewer mean population doublings (MPDs) than the controls prior to senescence. These lost divisions were mainly accounted for by slow cell turnover of the LTC cultures and by permanent cell cycle exit of 94% of the LTC cells, which resulted in many cell divisions being unmeasured by the MPD method. In the LTC cultures, p27(KIP1) accumulated and pRb became under-phosphorylated and under-expressed. Also, coincident with permanent cell cycle exit and before 1 MPD was completed, the LTC cultures upregulated the cell cycle inhibitors p21(WAF) and p16(INK4A) but not p14(ARF) and developed other markers of senescence. We then tested the relationship between cell cycle re-entry and the cell cycle-inhibitory proteins following subculture of the LTC cultures. In these cultures, the downregulation of p27(KIP1) and the phosphorylation of pRb preceded the complete resumption of normal proliferation rate, which was accompanied by the down-regulation of p16(INK4A). Our results show that most normal human fibroblasts can accumulate p16(INK4A), p21(WAF) and p27(KIP1) and senesce by cell division-independent mechanism(s). Furthermore, this form of senescence likely requires p16(INK4A) and perhaps p27(KIP1).

Detection of functional PTEN lipid phosphatase protein and enzyme activity in squamous cell carcinomas of the head and neck, despite loss of heterozygosity at this locus.

The human tumour suppressor gene PTEN located at 10q23 is mutated in a variety of tumour types particularly metastatic cases and in the germline of some individuals with Cowdens cancer predisposition syndrome. We have assessed the status of PTEN and associated pathways in cell lines derived from 19 squamous cell carcinomas of the head and neck. Loss of heterozygosity is evident at, or close to the PTEN gene in 5 cases, however there were no mutations in the remaining alleles. Furthermore by Western analysis PTEN protein levels are normal in all of these SCC-HN tumours and cell lines. To assess the possibility that PTEN may be inactivated by another mechanism, we characterized lipid phosphatase levels and from a specific PIP3 biochemical assay it is clear that PTEN is functionally active in all 19 human SCCs. Our data strongly suggest the possibility that a tumour suppressor gene associated with development of SCC-HN, other than PTEN, is located in this chromosomal region. This gene does not appear to be MXI-1, which has been implicated in some other human tumour types. PTEN is an important negative regulator of PI3Kinase, of which subunit alpha is frequently amplified in SCC-HN. To examine the possibility that PI3K is upregulated by amplification in this tumour set we assessed the phosphorylation status of Akt, a downstream target of PI3K. In all cases there is no detectable increase in Akt phosphorylation. Therefore there is no detectable defect in the PI3K pathway in SCC-HN suggesting that the reason for 3q26.3 over-representation may be due to genes other than PI3K110alpha.

Decreased expression of TGF-beta cell surface receptors during progression of human oral squamous cell carcinoma.

This study examined the immunocytochemical expression of the transforming growth factor-beta (TGF-beta) isoforms TGF-beta1, TGF-beta2, and TGF-beta3, together with the TGF-beta cell surface receptors TbetaR-I and TbetaR-II, in patient-matched tissue pairs of normal human oral epithelium, primary squamous cell carcinomas, and metastatic lymph node tumour deposits. There were no significant differences in the intensity of TGF-beta isoform specific staining between the normal oral epithelium, the primary tumours, and the lymph node metastases. By contrast, there was significantly less TbetaR-II in the metastases than in the primary tumour and between the primary tumour and the normal oral epithelium. Similar trends were evident with TbetaR-I, but not at a statistically significant level. This study also examined the structure of TbetaR-I and TbetaR-II in normal human oral keratinocytes in vitro and in 14 human oral carcinoma cell lines with known responses to TGF-beta1. No structural abnormalities of TbetaR-II were present in the normal keratinocytes or in 13 of 14 malignant cell lines; in one line, there were both normal and mutant forms of TbetaR-II, the latter being in the form of a frameshift mutation with the insertion of a single adenine base (bases 709-718, codons 125-128), predicting a truncated receptor having no kinase domain. No defects were present in TbetaR-I. The structures of TbetaR-I and TbetaR-II did not correlate with growth inhibition by TGF-beta1. The data suggest that decreased expression of TGF-beta receptors, rather than structural defects of these genes, may be important in oral epithelial tumour progression. In order to examine the functional significance of a specific decrease in TbetaR-II expression, a dominant-negative TbetaR-II construct (dnTbetaR-II) was transfected into a human oral carcinoma cell line with a normal TGF-beta receptor profile and known to be markedly inhibited by TGF-beta1. In those clones that overexpressed the dnTbetaR-II, growth inhibition and Smad binding activity were decreased, whilst the regulation of Fra-1 and collagenase-1 remained unchanged following treatment with TGF-beta1. The results demonstrate that a decrease in TbetaR-II relative to TbetaR-I leads to selective gene regulation with loss of growth inhibition but continued transcription of AP-1-dependent genes that are involved in the regulation of the extracellular matrix.

Replicative senescence as a barrier to human cancer.

There is evidence that one critically short telomere may be recognized as DNA damage and, as a consequence, induce a p53/p21WAF- and p16INK4A-dependent G1 cell cycle checkpoint to cause senescence. Additionally, senescence via a p53- and p16(INK4A)-dependent mechanism can be induced by the over- or under-stimulation of certain signalling pathways that are involved in cancer. Central to this alternative senescence mechanism is the p14ARF protein, which connects oncogene activation, but not DNA damage, to p53 activation and senescence. We find that immortal keratinocytes almost invariably have dysfunctional p53 and p16 and have high levels of telomerase, but very often express a wild-type p14(ARF). Furthermore, when normal keratinocytes senesce they show a striking elevation of p16 protein, but not of p14(ARF) or its downstream targets p53 and p21(WAF). These results suggest that p16, rather than p14(ARF), is the more important gene in human keratinocyte senescence, but do not exclude a co-operative role for p14(ARF), perhaps in the induction of senescence by activated oncogenes in neoplasia. Regardless of mechanism, these results suggest that replicative senescence acts as a barrier to human cancer development.

E-cadherin at the cell periphery is a determinant of keratinocyte differentiation in vitro.

The origin of the signal for keratinocyte differentiation is still unknown. Here, we show that Ca(2+)- and density-induced translocation of E-cadherin, but not P-cadherin, is accompanied by induction of differentiation-specific proteins in cultured keratinocytes. Antibodies that artificially cluster cell-surface E-cadherin in low extracellular Ca(2+) also induce differentiation-specific proteins, implicating E-cadherin as a determinant of keratinocyte differentiation in vitro.

The catalytic activity of the Src family kinases is required to disrupt cadherin-dependent cell-cell contacts.

Despite the importance of epithelial cell contacts in determining cell behavior, we still lack a detailed understanding of the assembly and disassembly of intercellular contacts. Here we examined the role of the catalytic activity of the Src family kinases at epithelial cell contacts in vitro. Like E- and P-cadherin, Ca(2+) treatment of normal and tumor-derived human keratinocytes resulted in c-Yes (and c-Src and Fyn), as well as their putative substrate p120(CTN), being recruited to cell-cell contacts. A tyrosine kinase inhibitor with selectivity against the Src family kinases, PD162531, and a dominant-inhibitory c-Src protein that interferes with the catalytic function of the endogenous Src kinases induced cell-cell contact and E-cadherin redistribution, even in low Ca(2+), which does not normally support stable cell-cell adhesion. Time-lapse microscopy demonstrated that Src kinase inhibition induced stabilization of transiently formed intercellular contacts in low Ca(2+). Furthermore, a combination of E- and P-cadherin-specific antibodies suppressed cell-cell contact, indicating cadherin involvement. As a consequence of contact stabilization, normal cells were unable to dissociate from an epithelial sheet formed at high density and repair a wound in vitro, although individual cells were still motile. Thus, cadherin-dependent contacts can be stabilized both by high Ca(2+) and by inhibiting Src activity in low (0.03 mM) Ca(2+) in vitro.

Increased dosage and amplification of the focal adhesion kinase gene in human cancer cells.

Focal adhesion kinase (pp125FAK) is present at sites of cell/extracellular matrix adhesion and has been implicated in the control of cell behaviour. In particular, as a key component of integrin-stimulated signal transduction pathways, pp125FAK is involved in cellular processes such as spreading, motility, growth and survival. In addition, a number of reports have indicated that pp125FAK may be up-regulated in human tumour cells of diverse origin, and consequently, a role has been proposed for pp125FAK in the development of invasive cancers. However, to date the mechanisms that lead to elevated pp125FAK expression in tumour cells have not been determined. Here we used in situ hybridization to confirm chromosome 8q as the genomic location of the human fak gene and report that elevation of pp125FAK protein in cell lines derived from invasive squamous cell carcinomas is accompanied by gains in copy number of the fak gene in all cases examined. In addition, we observed increased fak copy number in frozen sections of squamous cell carcinomas. Furthermore, increased dosage of the fak gene was also observed in many cell lines derived from human tumours of lung, breast and colon, including two cell lines Calu3 and HT29, in which fak was amplified. In addition, in an in vitro model for human colon cancer progression there was a copy number gain of the fak gene during conversion from adenoma to carcinoma, which was associated with increased pp125FAK protein expression. Thus, we show for the first time that many cell lines derived from invasive epithelial tumours have increased dosage of the fak gene, which may contribute to the elevated protein expression commonly observed. Although other genes near the fak locus are co-amplified or increased in copy number, including the proto-oncogene c-myc, the biological properties of pp125FAK in controlling the growth, survival and invasiveness of tumour cells, suggest that it may contribute to the selection pressure for maintaining increased dosage of the region of chromosome 8q that encodes these genes.

Role of the alternative INK4A proteins in human keratinocyte senescence: evidence for the specific inactivation of p16INK4A upon immortalization.

The INK4A locus on human chromosome 9p21 encodes two genes that have been implicated in replicative senescence and tumor suppression, p16INK4A and p14ARF. In contrast to p16INK4A, which is up-regulated to high levels, we were unable to detect p14ARF protein in senescent human keratinocytes. Also, p53, an established target of p14ARF, did not increase, suggesting that p14ARF is not instrumental in human keratinocyte senescence. In neoplastic keratinocyte cultures, p16INK4A inactivation was invariably associated with the immortal phenotype, and there was evidence for the inactivation of p16INK4A, independent of p14ARF, in 6 of 10 lines that lacked large homozygous deletions. In contrast, we failed to detect exon 1beta mutations or p16INK4A-independent deletions. These results emphasize the previously proposed role for p16INK4A in human keratinocyte senescence but do not rule out a supporting role for p14ARF inactivation.

Genetic and functional analyses exclude mortality factor 4 (MORF4) as a keratinocyte senescence gene.

Approximately 50% of immortal human keratinocyte lines show loss of heterozygosity of chromosome region 4q33-q34, and the reintroduction of chromosome 4 into one such line, BICR 6, causes proliferation arrest and features of replicative senescence. Recently, a candidate gene, mortality factor 4 (MORF4), was identified in this region and sequenced in 21 immortal keratinocyte lines. There were no mutations or deletions, and two of the seven lines that showed loss of heterozygosity at 4q33-q34 were heterozygous for MORF4 itself. Furthermore, the transfer of a chromosomal segment containing the entire MORF4 gene did not mimic the senescence effect of chromosome 4 in BICR 6. These results suggest that the inactivation of MORF4 is not required for human keratinocyte immortality.

Analysis of the CAVEOLIN-1 gene at human chromosome 7q31.1 in primary tumours and tumour-derived cell lines.

We identified CAVEOLIN-1 as a candidate for a tumour suppressor gene mapping to human chromosome 7q31.1. A number of studies suggest that caveolin could function as a tumour suppressor. Expression of caveolin, and in turn the number of caveolae within a cell, are inversely correlated with the transforming ability of numerous oncoproteins, including H-ras, v-abl, and bcr-abl, and caveolin is a major transformation-dependent substrate of v-src. Heterologous expression of caveolin has been shown to abrogate anchorage-independent growth and induce apoptosis in transformed fibroblasts and also to suppress anchorage-independent growth in human mammary carcinoma cells. We have analysed the status and expression of the human CAVEOLIN-1 gene in primary tumours and tumour-derived cell lines. We found no evidence for mutation of CAVEOLIN-1 in human cancers. Additionally, we found that while the first two exons of CAVEOLIN-1 are associated with a CpG island, this is not methylated in either primary tumours or in tumour-derived cell lines in which Caveolin-1 expression is low or undetectable. The level of expression of Caveolin-1 does not correlate with loss of heterozygosity at the CAVEOLIN-1 locus in these same cell lines. Contrary to other published studies, we have shown that CAVEOLIN-1 is not expressed in normal breast ductal epithelial cells in vivo. CAVEOLIN-1 is however highly expressed in breast myoepithelial cells and its expression is retained in tumours derived from breast myoepithelium. Together our data refute a role for CAVEOLIN-1 as a breast tumour suppressor gene in vivo.