The FOXM1 transcription factor is commonly overexpressed in a variety of human malignancies and several studies have suggested that this may be associated with malignant transformation [reviewed in 2]
. Despite the fact that the FOXM1 gene locus (12p13.3) has been reported to be amplified in HNSCC 
, its mechanism of action in oral carcinogenesis has not yet been investigated. This study provided the first evidence that FOXM1 is upregulated in oral dysplasias and HNSCC samples collected from various geographically distinct patient cohorts consisting of both Caucasians and Asians (,). Our bioinformatics analysis on published microarray data on various other types of human premalignant lesions and primary cancers confirmed that FOXM1 gene is commonly upregulated in a number of premalignant tissues supporting its ubiquitous role in early oncogenesis.
Tobacco (both smoke and smokeless forms) and betel quid use are the main risk factors for the development of HNSCC 
. At concentrations (0.01–10 mM) relevant to tobacco chewers, nicotine, but not arecoline and arecaidine was found to activate endogenous FOXM1 transcriptional activity through upregulation of mRNA, protein levels and increased protein stabilisation through phosphorylation in a dose-dependent fashion (). The co-carcinogenic effect of nicotine supports the finding that the risk of developing HNSCC is increased to 15 times higher in patients chewing betel quid with tobacco compared to chewing betel quid alone 
. Nicotine is currently perceived as a safe and non-carcinogenic substance often used in tobacco replacement therapy, despite increasing evidence that nicotine exhibit anti-apoptotic effect in oral epithelial cells 
. The growth promoting effect of nicotine may explain the effect of nicotine on FOXM1 activity. It has been shown that salivary nicotine concentration can reach millimolar concentrations in tobacco chewers. Nicotine chewing gum (eg. Nicorette, Nicotinell) contains between 2–4 mg nicotine and is expected to produce millimolar concentrations in the oral cavity. Some nicotine skin patches contain up to 21 mg nicotine (NicoDerm) which can deliver up to ~3 mM/hr (500 ng.hr/ml) of plasma nicotine level 
and is expected to produce much higher nicotine concentration locally on the skin. Although no evidence to date links nicotine exposure alone to human cancer, it has been previously shown that nicotine is co-carcinogenic with a tobacco smoke derivative benzo(a)pyrene in causing skin cancer in mice 
. A recently marketed nicotine-supplemented drink (NICLite®, Nichonica Ltd., California, USA) contains 4 mg nicotine (2 cigarettes worth) and will produce up to 0.8 mM (125 ng/ml) plasma nicotine level within 6 hrs. We have now shown that a concentration of nicotine as low as 0.1 µM was able to elevate FOXM1 protein levels in primary human oral keratinocytes (). More alarmingly, nicotine-laced alcoholic drinks (eg. Nicotini, a tobacco infused vodka or martini) are being offered in bars and restaurants following smoking bans in various developed countries. It has been reported that the synergistic actions of severe alcohol and tobacco abuse can have over 100-fold greater risk of developing HNSCC 
. Our findings that nicotine can directly activate the FOXM1 oncogene and the combined action of FOXM1B upregulation and nicotine exposure may induce malignant transformation () should prompt a reassessment of the widespread use of nicotine-based tobacco replacement therapies.
A model mechanism of FOXM1-induced oncogenesis.
We believe that FOXM1B overexpression (1st hit) alone in the premalignant SVpgC2a cells may eventually induce transformation if cells were subjected to prolonged cell culture to allow sufficient time to accumulate further oncogenic hits. We did not see spontaneous FOXM1B-induced transformation during the course of our study (6 months; data not shown). Hence, a semi-toxic dose of nicotine (10 mM – a concentration relevant to tobacco chewers) was used to expedite the selection of FOXM1B-transformed SVpgC2a cells. Whilst nicotine at lower doses (0.1–1 mM) activated FOXM1, nicotine at 10 mM did not activate FOXM1 activity due to toxicity (). This dose-related dual effect of nicotine may explain its co-carcinogenic activity, ie, at low doses it activates FOXM1B and at high doses it induces cell death thereby enhancing oncogenic selection. Because FOXM1B provides survival advantage, we believe that the fluctuating nicotine exposure in long-term tobacco users serves as a selection mechanism for oral cells with increased FOXM1B expression thereby increasing the risk of malignant transformation through subsequent genomic instability and accumulation of further oncogenic hits.
We have shown for the first time using genome-wide SNP profiling approach that acute upregulation of FOXM1B alone in primary normal human oral keratinocytes was able to induce random genomic instability in the form of CNA but not LOH (), indicating that FOXM1 overexpression alone is not sufficient to induce LOH formation in normal cells. Nevertheless, the upregulation of FOXM1B alone in a premalignant oral keratinocyte line was able to induce LOH involving the whole of chromosome 13. This is consistent with our previous finding that chromosome 13 (13q14–q33) was a hotspot LOH in a premalignant condition oral submucous fibrosis 
and others have shown that LOH in 13q21 is linked to the progression of HNSCC 
. The nicotine/FOXM1B-transformed malignant cell clones (SVFN1-8; & ) exhibited multiple additional LOH/CNA loci. The non-random pattern of LOH/CNA loci in these malignant clones supports the phenomenon of field cancerisation often seen in HNSCC 
and numerous other cancers such as lung, oesophagus, vulva, cervix, colon, breast, bladder and skin 
Although FOXM1B at physiological levels has been reported as a regulator of DNA repair 
, its upregulation is likely to interfere with the normal DNA repair mechanism leading to enhanced genomic instability rather than enhanced DNA repair. It is known that levels of many cell cycle-related genes are tightly regulated and any deviations from their physiological critical concentrations may cause problems. It has been shown that silencing FOXM1 leads to genomic instability via aberrant expression of mitotic spindle assembly genes such as CEPN-F, Aurora B and Plk1, which are putative targets of FOXM1 
. We have shown that overexpression of FOXM1B can also cause genomic instability. In support, numerous studies have demonstrated that proteins which are important in mitotic division are often found amplified in human cancers giving rise to genomic instability 
. Whilst DNA repair and genomic instability may, or may not, be directly inter-linked, FOXM1B induction may trigger both DNA repair (aberrant repair) as well as genomic instability. It is important to emphasise that our data () showed that FOXM1 gene is significantly upregulated in genomically unstable oral premalignant (dysplasia and erythroplakia) and HNSCC samples 
. Moreover, upregulation of FOXM1B in a wide variety of human cancers 
is consistent with our data that upregulation of FOXM1B causes genomic instability. The findings that FOXM1B induces DNA repair genes such as XRCC1 and BRCA2 
and the fact that most cancers contain genomic instability 
suggest that activation of DNA repair and genomic instability may not be mutually exclusive in cancer cells. Furthermore, the upregulation of FOXM1B in majority of human cancers suggests that gain of FOXM1B function is an important step in oncogenesis.
In search of FOXM1B-targeted malignancy associated genes, micro-regions of consensus LOH loci shared amongst the 8 malignant clones were compared. The most common recurrent LOH loci were found in 4p, 10q and 18q. Whilst this study did not identify candidate genes within the 4p and 18q, these two loci have been previously linked to HNSCC progression 
. Our previous study showed that LOH in 10q22–24 and 18q21 were hotspots in oral premalignancy 
and others have linked 10q with poor HNSCC prognosis 
and 18q with metastasis of HNSCC 
. Within the amplified 10q23.33 LOH locus, we have identified a centrosomal protein CEP55 (also known as URCC6, C10orf3 or FLJ10540) and a stem cell marker/DNA helicase/chromatin remodelling factor HELLS (also known as LSH, PASG, SMARCA6, FLJ10339 or Nbla10143) 
to be likely mediators of FOXM1B-induced malignant transformation. CEP55 and HELLS gene amplifications were found to be the commonest among the transformed keratinocyte clones. In support, the endogenous mRNA of both CEP55 and HELLS were significantly activated by constitutive expression of FOXM1B in primary human normal oral keratinocytes. Consistently, bioinformatics analysis of published microarray data in FOXM1B-knockdown (by siRNA) breast cancer cells 
showed significantly decreased levels of CEP55 (Supplemental Fig. S5
). Our ChIP-qPCR results suggest that FOXM1 protein binds directly to the promoter of CEP55 () but not HELLS () indicating that CEP55 is a direct transcriptional target of FOXM1, whilst HELLS may be an indirect target. This finding may explain the lower correlation between HELLS and FOXM1 expression ( and Supplemental Fig. S6E
The upregulation of CEP55 and HELLS by FOXM1B in the absence of nicotine suggests that nicotine acts as an upstream enhancer of FOXM1B activity. Nicotine alone may not be sufficient to achieve a critical FOXM1B threshold level required for malignant transformation. Consistent with this notion, nicotine alone failed to upregulate CEP55 or HELLS mRNA expression (data not shown). This implies that nicotine exerts its co-carcinogenic properties primarily through the enhancement of FOXM1B rather than direct activation of CEP55 or HELLS. Since FOXM1B alone can activate de novo
expression of CEP55 and HELLS, we hypothesised that this de novo
activation constitute an initial oncogenic stress (1st
hit) which subsequently led to a secondary genomic instability event resulting in a sustained CEP55 and HELLS expression through copy number amplification of 10q23 loci (2nd
hit). According to this model (), at later stages of malignancy, the expression of CEP55 and HELLS may or may not be dependent on FOXM1B. In support for this model, it has been shown previously 
that upregulation of cyclin D1 (CCND1) expression was detected in oral hyperplasia/premalignant lesions and CCND1 gene locus (11q13) amplification was found in early dysplasia, carcinoma in situ and also metastatic HNSCC. Furthermore, upregulation of CCND1 gene expression was detected in premalignant lesions adjacent to HNSCC tumours with amplified gene locus. In contrast no upregulation of CCND1 gene expression was detected in the premalignant lesions adjacent to non-amplified tumours. Based on this model, we therefore hypothesised that upregulation of CEP55 and HELLS precedes gene amplification during malignant transformation induced by FOXM1B.
Both CEP55 and HELLS mRNA levels were found to be significantly elevated in both oral premalignant and HNSCC cell lines compared to normal oral keratinocytes (). Consistent with their roles in malignancy, CEP55 is upregulated in hepatocellular carcinoma and its overexpression induces anchorage-independent growth, enhanced cell growth at low serum levels and induction of tumourigenesis in nude mice 
. FOXM1 is known to be overexpressed in hepatocellular carcinoma 
. A recent study involving 3 independent breast cancer microarray datasets containing a total of 699 patients revealed that CEP55 and FOXM1 are amongst the signature prognostic markers which predicts breast cancer outcome 
. HELLS has been implicated in leukaemia 
, non-small cell lung cancer 
, breast cancer 
, and melanoma 
. HELLS was recently identified as one of the consensus genes expressed in human embryonic stem cells 
. The role of CEP55 in midbody/cytokinesis 
and HELLS in epigenetic DNA methylation and chromatin maintenance 
further support the role of FOXM1B in genome maintenance 
. We hypothesise that aberrant upregulation of FOXM1B may be inducing genomic instability through a program of malignant transformation involving deregulation of cytokinesis (CEP55) and epigenetic modifications (HELLS) whereby genomically unstable cells acquire oncogenic survival advantage ().
In summary, this study provides, for the first time, several lines of evidence that FOXM1B is upregulated early during oral pre-malignancy and its expression persists in primary and metastatic HNSCC tissues. Cellular, molecular and genomic evidence have shown that nicotine, at concentrations relevant to tobacco chewers, significantly activated FOXM1 transcriptional activity in oral keratinocytes. The synergistic action of nicotine and FOXM1B overexpression led to anchorage-independent malignant transformation of a premalignant oral keratinocyte line which exhibited non-random genomic instability LOH/CNA profiles. Analysis of consensus LOH loci identified CEP55 and HELLS FOXM1B-target genes for malignant transformation. In conclusion, the potential of FOXM1B, CEP55 and HELLS as early oral cancer biomarkers deserves further investigation.