Identification of novel tumour-associated genes differentially expressed in the process of squamous cell cancer development


Chemically induced mouse skin carcinogenesis represents the most extensively utilized animal model to unravel the multistage nature of tumour development and to design novel therapeutic concepts of human epithelial neoplasia. We combined this tumour model with comprehensive gene expression analysis and could identify a large set of novel tumour-associated genes that have not been associated with epithelial skin cancer development yet. Expression data of selected genes were confirmed by semiquantitative and quantitative RT-PCR as well as in situ hybridization and immunofluorescence analysis on mouse tumour sections. Enhanced expression of genes identified in our screen was also demonstrated in mouse keratinocyte cell lines that form tumours in vivo. Self-organizing map clustering was performed to identify different kinetics of gene expression and coregulation during skin cancer progression. Detailed analysis of differential expressed genes according to their functional annotation confirmed the involvement of several biological processes, such as regulation of cell cycle, apoptosis, extracellular proteolysis and cell adhesion, during skin malignancy. Finally, we detected high transcript levels of ANXA1, LCN2 and S100A8 as well as reduced levels for NDR2 protein in human skin tumour specimens demonstrating that tumour-associated genes identified in the chemically induced tumour model might be of great relevance for the understanding of human epithelial malignancies as well.


Nonmelanoma skin cancer, such as basal cell carcinoma (BCC) and squamous cell carcinoma (SCC) is the most common malignancy in the Caucasian population (Dooley et al., 2003). Both subtypes occur primarily on sun-exposed areas of the body and have been strongly associated with chronic sun exposure (Bowden, 2004). SCC as a solid tumour is composed of transformed epidermal keratinocytes with a highly invasive growth and tendency to metastasize. The cytogenetic pattern of cutaneous SCC is irrespective of the initiating event (tobacco, coal, UV) and very similar to SCC at other anatomic sites like head and neck SCC, suggesting that common pathways may play an important role in development of different types of SCC. It was demonstrated by in vitro and in vivo model systems that malignant transformation of epidermal cells is a multistage process, in which stepwise accumulation of genetic and epigenetic events determines the transition from normal to malignant cellular state. However, the onset and the order of genetic alterations that lead to development of most sporadic cancers remain undefined.

For the past half century, mouse skin carcinogenesis has been an important tool for developing the current concepts regarding human neoplasia and the multistage nature of tumour development (Lu et al., 2001). In recent years, it was demonstrated, that some types of mutation in oncogenes and tumour suppressor genes identified in mouse skin models also occur in human epithelial cancers. One of the best-defined experimental in vivo systems for epithelial cancer development is the chemically induced tumour model of mouse back skin (Marks and Furstenberger, 1990; DiGiovanni, 1992; Yuspa, 1998). Here, treatment of the skin with the carcinogen 7,12-dimethylbenz-[α]-anthracene (DMBA) and the tumour promoter 12-O-tetradecanoylphorbol-13-acetate (TPA) results in the formation of benign papillomas (PAPs) and malignant tumours (SCCs). Using the model of chemical-induced skin carcinogenesis, the timing of genetic and chromosomal alterations that take place during the different stages of tumour development can be studied (DiGiovanni, 1992; Yuspa, 1998). Recently, we have identified TPA-inducible genes in mouse back skin using suppression subtractive hybridization and gene expression profiling (Breitenbach et al., 2001; Gebhardt et al., 2002, 2005; Schlingemann et al., 2003). Some of the identified TPA-inducible genes also revealed an enhanced expression in advanced tumour stages, suggesting a potential contribution to tumour progression and metastasis. Comprehensive and systematic high-throughput comparison of gene expression profiles has not been possible until the advent of functional genomics. cDNA microarrays are an ideal approach for systematic comparison, because they simultaneously measure the expression of a large number of genes. They have been increasingly used to identify biomarkers of tissues and potential molecular targets for anticancer therapy (Clarke et al., 2001). Although there are many reports on the expression profiling of head and neck, oesophageal, oral or cutaneous SCC, none has compared the alteration of the network of tumour-associated genes in the process of cancer development. In this study, we analysed specimens of short-term TPA-treated back skin as well as benign and malignant tumours derived from the chemical-induced carcinogenesis model using two independent cDNA microarrays. The combination of chemical-induced carcinogenesis and global gene expression profiling revealed differential expression of numerous genes with well-known deregulation during epithelial tumorigenesis as well as a large set of cDNAs representing known and unknown genes that have not yet been described in the context of epithelial tumour development. Functional annotation allowed classification of these genes according to their association with critical tumour-associated processes, such as cell cycle regulation, apoptosis, tissue remodelling and cell adhesion, during mouse and human skin malignancy.


Identification of differentially expressed genes during multistage skin carcinogenesis

In an effort to identify novel genes that are expressed during the establishment and progression of epithelial cancer, we combined the in vivo model of chemical-induced skin cancer with comprehensive gene expression profiling. Pools of mRNA samples derived from TPA-treated back skin, benign (PAP), or malignant (SCC) skin tumours were hybridized together with mRNA from skin of age-matched controls on two independent sets of cDNA microarrays – the 20k ArrayTAG™ (LION Bioscience) and the 15k NIA (National Institute of Aging) chip (Supplementary Figure 1; see Materials and methods for details). In order to correct for dye-specific labelling and fluorescence intensity, each sample was analysed in colour switch experiments where the assignment of fluorochromes was reversed. Additionally, all samples were evaluated in an independent replica experiment. Finally, the data from all experiments were integrated and only those data were considered that revealed a significant differential expression (corresponding to a onefold change on a ln scale or a 2.71-fold change on a linear scale) and were confirmed in repetitive experiments. To demonstrate the quality and reproducibility of our data, we initially compared differential expression of genes present on both cDNA microarrays. For annotated genes, which revealed a significant altered expression in SCC, we could illustrate a high correlation (R2=0.78) between the ratios derived from the microarray of the NIA and the ArrayTAG™ microarray (Supplementary Figure 2). Similar correlations between both arrays were also observed for expression in TPA-treated back skin and PAP (data not shown). The small number of genes with opposite regulation on individual chips was not considered for further analysis. After filtering and normalization, we identified a total of 6543 expressed, nonredundant annotated cDNA clones with 1495 present on both arrays. From the 6543 genes 1426 showed a significantly altered gene expression (greater than 2.71-fold) in at least one of the three analysed conditions and out of this list genes with the strongest up- or downregulation in benign or malignant tumours were listed in Table 1. cDNA clones present on the two types of cDNA microarrays together with the differentially expressed genes in each stage are available as supplementary data at

Table 1 Summary of genes with highest up- or downregulation in (a) benign tumours and (b) malignant tumours

Independent validation of microarray data

Differential expression of selected genes present on the ArrayTAG™, the NIA, or both arrays was confirmed by semiquantitative reverse transcription PCR (RT-PCR, Figure 1a and Supplementary Figure 3) and altered mRNA levels in SCC were approved for Stmn1 (6.7-fold repressed), Ecm1, Crabp2 and Tnrfs12a (5.6-, 3.6- and 3.2-fold induced, respectively) by real-time quantitative reverse transcription PCR (RQ-PCR). To proof for a positive correlation between elevated expression and neoplastic transformation of keratinocytes, we used a collection of mouse keratinocyte cell lines that differ in their tumorigenesis after subcutaneous injection in nude mice. Corresponding to high expression in advanced tumour samples, eight selected genes (Lcn2, Fosl1, Adam8, Anxa1, Nr6a1, 2210418J09Rik, 4930579A11Rik and 56303400A09Rik) revealed a significant upregulation in established cell lines characterized by a benign or malignant phenotype compared to nontumorigenic cell lines (Figure 1b).

Figure 1

Verification of microarray data by semiquantitative RT-PCR analysis. (a) RT-PCR was performed with RNA of short-term TPA-induced dorsal skin (lane 2), PAP (lane 4) and SCC (lane 6) together with control skin of age- and sex-matched animals (lanes 1, 3 and 5). The expression of these genes is illustrated that are constitutively upregulated (Lcn2, Fosl1, Adam8, Anxa1 and 2210418J09Rik), downregulated in advanced tumour stages (Gsn, Ndr2 and Fhl1) or specifically upregulated in distinct stages of tumourigenesis (Nr6a1, 4930579A11Rik and 56303400A09Rik). Tubb5 served as a control for cDNA quality and quantity. (b) Expression of tumour-associated genes in keratinocyte cell lines was measured by RT-PCR with RNA of nontumorigenic keratinocyte cell lines (lanes 1–2), cells forming benign (lanes 3–4) or malignant (lanes 5–9) tumours in vivo demonstrating a positive correlation between enhanced expression of Lcn2, Fosl1, Adam8, Anxa1, Nr6a1, 2210418J09Rik, 4930579A11Rik and 56303400A09Rik in neoplastic transformation. Tubb5 served as control for cDNA quality and quantity.

Finally, we performed in situ hybridization (ISH) analysis on tissue sections derived from chemical-induced tumours, in order to determine the cellular compartments that abundantly express the mRNA of selected genes. Expression of known genes, such as Anxa1, Lcn2 and Tnfrsf12a, but also of the unknown gene 2210418J09Rik could be observed in neoplastic keratinocytes of both PAP and SCC (Figure 2). In contrast, expression was significantly weaker or even below the level of detection in stromal tissue or areas of ‘normal’ skin adjacent to the tumour tissue (data not shown). Moreover, we could confirm loss of expression in benign and malignant tumours for downregulated genes, such as Ndr2 and S100a3, which were both expressed in keratinocytes of normal skin (NS) (Figure 3; data not shown).

Figure 2

Expression of induced genes in advanced stages of mouse tumours. The analysis was performed on 6 μm paraffin sections derived from chemically induced tumours. Sections of papillomas (ah) and squamous cell carcinomas (il) were hybridized with 35S-labelled antisense riboprobes derived from specific cDNA-fragments of Anxa1 (e and i), Lcn2 (f and j), Tnfrsf12a (g and k) and 2210418J09Rik (h and l). Following counterstaining with eosin and haematoxylin, pictures were photographed under bright field conditions. All investigated genes exhibit expression in neoplastic keratinocytes of papillomas (eh) and SCCs (il), which is illustrated by a black signal. 35S-labelled sense riboprobes served as control for specificity (ad). Scale bar=50 μm.

Figure 3

Impaired expression of Ndr2 protein in mouse skin tumours. Reduced Ndr2 protein level was confirmed in mouse skin carcinogenesis by immunofluorescence analysis. Ndr2 protein was detectable in keratinocytes of normal skin (a, red signal), but not in SCC (b) or normal skin incubated without the anti-Ndr2-specific antibody (c). H33342 was used for nuclear staining (blue signal). Scale bar=50 μm.

Clustering analysis and functional classification of coexpressed genes

The set of genes with altered expression (≥2.8-fold) was used for self-organizing map (SOM) analysis to illustrate coregulation of genes in different stages of tumour development (Tamayo et al., 1999; Li and Johnson, 2002). We used a 4 × 3 matrix resulting in 12 characteristic profiles (Figure 4; supplementary data at with clusters 1.1–4 and 2.2–4 exhibiting repressed genes, whereas clusters 2.1 and 3.1–4 represent upregulated genes. Profiles of coexpressed genes could be further subdivided in (i) clusters with constant up- or downregulation in all analysed conditions (clusters 1.3, 1.4 and 3.1), (ii) clusters with altered expression only in short-term TPA treatment or benign tumours (clusters 2.1, 2.3, 2.4 and 3.3), (iii) clusters with TPA-independent expression and decreased expression in benign and malignant tumours (clusters 1.1, 1.2 and 2.2) and (iv) clusters with increased expression during tumour progression (clusters 3.2 and 3.4).

Figure 4

SOM nonhierarchical clustering analysis of gene expression at different stages of cutaneous SCC formation. A 4 × 3 matrix was used to illustrate coexpression of genes during chemically induced tumour development. The number of genes in each cluster is indicated (n). The gene lists of all clusters are available as online supplementary data ( Each cluster is represented by the centroid (average pattern) of genes together with the s.d. of each stage. For each graph the x-axis represents the analysed conditions (TPA, PAP, SCC), whereas the y-axis signifies the logarithmic (ln) expression values.

In order to elucidate biological processes that significantly correlate with epidermal malignancy, we analysed our sets of genes with altered expression in TPA-treated skin, PAP or SCC according to their functional annotation using the EASE software package (Table 2). These data were compared with functional categories obtained with the SOM cluster data sets. As expected, TPA-induced genes could be classified into defence, immune, stress and wound response accompanied by genes linked to regulation of signalling cascades. Most of the TPA-induced genes in these categories were only transiently elevated and, therefore, could be found in cluster 2.1. For genes with enhanced expression in benign tumours, we observed a significant enrichment of genes in the categories of apoptosis induction and regulation together with processes of catabolism. In contrast, some of the repressed genes belong to categories of fatty acid metabolism and energy reserve metabolism suggesting a major change in energy pathways at this stage of tumour development. Whereas, significant changes in apoptosis regulation were specifically detected in PAPs (cluster 3.3), alterations in catabolism and metabolism persist and were also present in malignant tumours. Our data also confirm the crucial role of cell cycle activation, extracellular proteolysis and integrin-mediated signalling pathways for multistage skin malignancy, particularly with regard to malignant progression (cluster 3.2).

Table 2 Functional annotation of genes present in individual SOM clusters according to the GO database

Expression of novel tumour-associated genes in human epithelial skin tumours

Finally, we asked whether principles of altered gene expression identified in the mouse model could be applied to the process of human skin carcinogenesis. Therefore, expression of selected tumour-associated genes was measured by ISH on tissue sections of human skin tumour specimen. In agreement with enhanced levels in the chemical-induced mouse tumours, strong expression of ANXA1, LCN2 and S100A8 mRNA was observed in neoplastic keratinocytes within the tumour tissue but not in ‘normal’ skin adjacent to the tumour area (Figure 5). We could also demonstrate loss of NDR2 protein expression in SCC compared to NS (Figure 6), which is in line with the observed downregulation in chemical-induced tumours of mouse back skin.

Figure 5

Induced expression of tumour-associated genes in human skin tumours. In situ hybridization on human SCCs was performed for the expression of ANXA1, LCN2 and S100A8. Sections of tumour specimen, characterized as well-differentiated SCCs, were hybridized with 35S-labelled antisense riboprobes derived from corresponding cDNA fragments. Following counterstaining with haematoxylin and eosin, sections were photographed under bright field conditions. Expression of genes was detectable in keratinocytes within the tumour area (df) but not in the adjacent ‘normal’ skin (ac). 35S-labelled sense riboprobes were used as control for specificity (gi). Scale bar=50 μm.

Figure 6

Reduced expression of NDR2 protein in human skin tumours. NDR2 protein expression was studied by immunofluorescence analysis and was detectable in keratinocytes of normal skin (a, red signal), but not in SCC (b) or normal skin incubated without the anti-NDR2-specific antibody (c). H33342 was used for nuclear staining (blue signal). Scale bar=50 μm.


The objective of this study was to define characteristic changes in global gene expression associated with epithelial tumour development. Although alterations in gene expression in the course of tumour promotion and progression have been considered to be critical for carcinogenesis in human, its significance could not be documented directly from human cancers but relied on animal systems, such as the mouse skin model of chemically induced multistage carcinogenesis (Ito et al., 1995). Here, we have employed this well-established experimental animal model and large-scale gene expression profiling to identify novel tumour-associated genes. It is well known that different types of genetic alterations can occur during multistage carcinogenesis and that these variations together with epigenetic changes are the main cause for differential expression of tumour-associated genes. Since our aim was the identification of the most common genes with altered levels in the defined stages of tumour development, we pooled for our study a multitude of individual samples to have a valid average over the heterogeneity in genetic changes. Although our strategy is limited by the fact that similar changes for differentially expressed genes could represent independent events in distinct tumours, this limitation does not take away the potential importance of the identified genes for the process of tumour promotion and progression.

The combination of the mouse tumour model and global gene expression analysis, together with stringent statistical and data filtering criteria, resulted in a comprehensive list of differentially expressed genes. In addition to the identification of genes (e.g., several cytokeratins as well as members of the MMP and integrin families) that are known to be differentially expressed and functionally correlated to epidermal cancer development (Rundhaug et al., 1997; Egeblad and Werb, 2002; Kerkela and Saarialho-Kere, 2003), several experimental settings demonstrated the power of our microarray approach. First of all, RNA of specimens were applied on two independent cDNA microarrays resulting in similar expression data in the great majority of genes present on both chips. Moreover, semiquantitative and real-time PCR as well as ISH and immunofluorescence analysis for selected genes were described in this study, but also in previous manuscripts, confirming differential expression (Gebhardt et al., 2002; Schlingemann et al., 2003). In accordance with our initial goal to identify novel tumour-associated genes, our list comprises known genes and unknown EST and Riken clones with altered mRNA patterns that have not been correlated to epidermal tumour development so far. These candidates provide a valuable source for further functional investigations to increase our knowledge on the principles of neoplastic transformation of epithelial tissues. In fact, we found elevated mRNA levels for some upregulated genes in mouse keratinocyte cell lines known to form benign or malignant tumours in nude mice, but not in nontumorigenic counterparts, supporting the idea that they might be involved in neoplastic transformation and tumour development in vivo. In line with this notion, recent studies have shown that the AP-1 family member Fra-1 (Fosl1) is essential for efficient neoplastic transformation of keratinocytes (Young et al., 2002), whereas expression of the membrane-bound metalloproteinase Adam8 was recently correlated with lung cancer (Ishikawa et al., 2004).

Coregulation of genes exposed by the SOM algorithm provides a basis for the identification of new concepts of transcriptional regulation and important transcription factors necessary for tumour promotion and progression into malignant tumour stages. Studies in recent years yielded important insights into the pathways and transcription factors that control the rate-limiting steps in multistage skin carcinogenesis. As an example, in vitro and in vivo model systems revealed a pivotal role of AP-1 in skin cancer induction as well as progression into malignant stages (Young et al., 2003). However, it is likely that the in vivo specificity of cellular gene activation is greatly influenced by combinatorial protein–protein interactions with other promoter-bound factors. Detailed analysis of regulatory elements within the promoter and enhancer regions of genes present in our cluster will not only result in the identification of novel target genes, but will also broaden our current understanding of the role of AP-1 and other transcription factors in tumorigenesis. Indeed, numerous genes (e.g., Fosl1, Itga5, Mmp9, Mmp13 and TnC) present in clusters 3.1 and 3.2 are well known AP-1 target genes (Angel et al., 2001) or share at least one potential AP-1-binding site in their proximal enhancer region (Hess J, unpublished data). It will be interesting to investigate, whether some of these genes are novel AP-1 target genes and whether they are causally involved in oncogenic transformation mediated by Fos and Jun proteins.

It has been postulated that coregulated genes exhibit similar functions and are often arranged in the same signalling pathway or functional network. Therefore, SOM analysis combined with functional classification further provides the possibility to gain information on functional concepts of multistage epithelial cancer development. As expected, TPA-induced genes could be classified into defence- and immune response as well as regulation of signal transduction control that is in agreement with the well-known induction of hyperplasia and inflammation in response to TPA treatment on mouse back skin. In benign tumours, we observed a transient upregulation of genes involved in induction and regulation of apoptosis. Although this observation opposed the current concept of apoptosis in preventing the survival of malignant cells, our data comply with a positive correlation between tumour progression and the presence of apoptotic cells that has been found in the process of squamous cell carcinogenesis (Stern et al., 1997). It has been postulated that epithelial cells in skin tumours, generated by chemical carcinogenesis, maintain the capacity to undergo apoptosis and that the presence of apoptotic cells in advanced PAPs could be an indication of endogenous mutagenic activity and high genomic instability. Thus, we conclude that in our chemical-induced SCC proliferation and expansion, rather than apoptosis inhibition, mainly contribute to malignancy. According to this concept, we found an obvious increase of genes associated with cell cycle regulation during tumour progression. However, we cannot formally exclude the possibility that induction of the apoptotic pathway is causally involved in the low potential of the majority of PAPs to convert into carcinomas. Accordingly, one could also speculate that apoptosis is suppressed in those few PAPs that have a high tendency to progress into malignant tumours.

Our data also confirm the important role of altered cell adhesion, integrin-mediated signalling and extracellular proteolysis for epidermal tumour formation (Arias, 2001; Cavallaro et al., 2002; Bogenrieder and Herlyn, 2003). Some of the induced proteases, such as the metalloproteinase Mmp9, Mmp13 and Mmp14, are well accepted to enforce invasive growth, metastasis and tumour angiogenesis and it will be a challenge for the future to elucidate the potential role of the membrane-bound metalloproteinase Adam8, which has been found to be coregulated, in these processes. In this regard, it is worthwhile to mention that Adam8 has been discussed to be a useful diagnostic marker for lung cancer and probably a therapeutic target (Ishikawa et al., 2004).

Finally, enhanced mRNA levels of tumour-associated genes, such as ANXA1, LCN2, S100A8 and impaired expression of NDR2 in human SCC tissue sections support our idea that individual genes but also clusters of coregulated genes identified in the experimental model of mouse skin carcinogenesis have a great relevance to neoplastic transformation of human epithelial cells. In line with this assumption, enhanced levels of ANXA1 were detected in lung and oral SCCs (Leethanakul et al., 2000; Petroziello et al., 2004) and LCN2 was highly expressed in hepatocellular carcinomas (Patil et al., 2005). Moreover, recent analysis of cancer-related genes with increased mRNA abundance in common human malignancies revealed significant transcriptional activation on human chromosome 1q21, also known as epidermal differentiation cluster (Glinsky et al., 2003a, 2003b). Many genes identified by our approach, such as members of the S100 and SPRR protein families, represent the mouse homologues of human genes that are clustered on 1q21. In contrast, analysis of oesophageal SCC using cDNA-microarray technology demonstrated a significant downregulation of S100 and SPRR family members (Luo et al., 2004), whereas altered expression of cell-adhesion molecules, MMPs and some oncogenes, such as Ccnd1, Myc and Fra-1, are in accordance with our data (Hu et al., 2001a, 2001b; Nair et al., 2005). These findings argue for some common alterations, but also obvious differences in the pathogenesis of oesophageal SCC and carcinogenesis of the skin. Altered expression of integrins and MMPs is also a hallmark for other SCCs derived from oral (Mendez et al., 2002; Nagata et al., 2003; Tsai et al., 2004), hypopharyngeal (Cromer et al., 2004), lung (Gogali et al., 2004; Kettunen et al., 2004) and cervical epithelium (Chen et al., 2003).

Future confirmation at the protein level and functional analysis of the identified genes in tissue culture and animal model systems will be required to define the individual roles that the identified genes play in tumour initiation, promotion and progression.

Materials and methods

Animal work and sample preparation

All procedures for performing animal experiments were carried out concordant to the principles and guidelines of the ATBW (officials for animal welfare). The dorsal skin of C57BL/6 and NMRI mice was treated with TPA as described (Breitenbach et al., 2001). Skin tumours derived from female NMRI mice were generated according to the initiation–promotion protocol of chemically induced multistage carcinogenesis (Furstenberger and Kopp-Schneider, 1995). Total RNA was isolated from acetone-treated control skin, 6 h TPA-treated dorsal skin, persistent PAP prepared 2 weeks following the last TPA application, SCC and the corresponding age- and sex-matched controls (Co_PAP and Co_SCC). Skin biopsies from patients with well-differentiated SCC were obtained from surgical excisions of the affected areas at the Department of Dermatology, University of Cologne. The patients signed the informed consent from the Department of Dermatology, University of Cologne, approved by the Institutional Commission of Ethics (Az. 9645/96). Procedures of mRNA isolation and labelling were described previously (Schlingemann et al., 2003).

cDNA microarrays

RNA from back skin of three different control- or TPA-treated animals was pooled in order to correct for individual gene expression variations. A mixture of pooled tumour lesions of DMBA/TPA-treated back skin derived from three independent animals (3 PAPs per mouse and 2 carcinomas per mouse) was used for RNA preparation to have a good coverage of the different types of genetic alteration that occur during multistage skin carcinogenesis. mRNA samples were hybridized on two different types of microarrays comprising mouse gene-specific cDNA fragments, (i) 20k ArrayTAG™ collection (LION Bioscience) and (ii) 15k cDNA clone set from the NIA kindly provided by Steve Scherer (Toronto). The 15k microarray contained 15 267 ‘unique’ cDNA clones that were derived from pre- and peri-implantation embryos, E12.5 female gonad/mesonephros and newborn ovary. Up to 50% of these clones were originated from novel genes expressed during mouse embryogenesis with an average insert size of 1.5 kb (Ko et al., 2000). In addition to the genes that were uniquely expressed in early mouse development, nearly 7500 show significant homology to known genes (Kargul et al., 2001). The 20k ArrayTAG™ included 20 172 sequence-verified cDNA clones (comprising about 10 000 annotated genes) with an average insert size of 200–600 bp. All clones were proved for the absence of repetitive elements and low-complexity regions. For each clone, PCR amplification was performed as described in Tanaka et al. (2000) and Schlingemann et al. (2003). After purity and quantity analysis by electrophoresis, the PCR-generated DNA fragments were precipitated and dissolved in spotting buffer containing 3 × SSC and 1.5 M betaine (Diehl et al., 2001). The clones were spotted on Nexterion™ Slide E (Schott Nexterion) together with positive control spots from the Spot Report System™ (Stratagene) as well as a variety of additional negative controls, for example, mouse C0t-1 DNA (Invitrogen), yeast tRNA (Sigma-Aldrich) and poly-dA (Amersham Biosciences). Furthermore, Cy3- and Cy5-labelled DNA was included to yield specific fluorescent landmarks on the arrays. The further procedures of microarray processing and hybridization were the same as described in Wrobel et al. (2003).

Data acquisition, preprocessing and analysis

All data sets for spots not recognized by the GenePixPro analysis software were excluded from further considerations. The remaining data sets were ranked according to spot homogeneity (as assayed by the ratio of median and mean fluorescence intensities), ratio of spot-to-local-background intensity and the variance of the logarithmic ratios of replicates. For each hybridization the intensities were normalized by variance stabilization (Huber et al., 2002) to reach a homogenous distribution of the variance for all intensity values of an experimental series. Accurate differential expression values for each gene were obtained by calculating the average of normalized ratios of replicate experiments after reversing the values obtained in the respective colour switch experiment. Data sets representing differentially expressed genes were selected applying the following criteria: (i) valuable data must be obtained in at least two independent experiments, (ii) mean intensities range significantly (at least 10-fold) above background and (iii) mean of normalized natural logarithmic ratios (ln-ratio) higher/lower than ±1 (meaning differential expression of more than ±2.71-fold in a linear scale).

Validation of microarray data

ISH and immunofluorescence analysis were performed with paraformaldehyde-fixed and paraffin-embedded sections of mouse and human skin tumours (6 μm) as described previously (Breitenbach et al., 2001; Schlingemann et al., 2003; Gebhardt et al., 2005). The polyclonal goat-anti-NDRG2 (E20) antibody (sc 19468, Santa Cruz) was used to detect Ndr2 protein on tumour sections by immunofluorescence analysis. The same RNA samples used for chip hybridization were also applied to semiquantitative RT-PCR following the protocols as described previously (Schlingemann et al., 2003). The mouse keratinocyte cell lines with different tumorigenic potential in vivo are described in Strickland et al. (1988), Krieg et al. (1991), Diaz-Guerra et al. (1992) and Rennecke et al. (1999). Oligonucleotide sequences and sequences of ISH probes are available upon request.

Clustering of coregulated genes and functional classification

Clustering via SOM was conducted using GeneSpring® software version 6.1 (Silicon Genetics). SOM is a nonhierarchical clustering technique similar to k-means clustering, which, in addition to dividing genes into groups based on expression patterns, illustrates the relationship between groups by arranging them in a two-dimensional map. Here, we applied a 4 × 3-SOM matrix with a radius of 1.0 and an iteration number of 100 000. The gene lists of each cluster are available as supplementary data at EASE analysis ( was either performed with sets of genes showing induced or repressed expression levels in TPA-treated skin, PAPs or SCC, or with sets of genes derived from distinct SOM clusters. Most significant categories (biological process) and corresponding genes within the SOM clusters are available as supplementary data at



basal cell carcinoma




in situ hybridization


National Institute of Aging


normal skin




real-time quantitative reverse transcription PCR


reverse transcription PCR


squamous cell carcinoma


self-organizing maps




  1. Angel P, Szabowski A, Schorpp-Kistner M . (2001). Oncogene 20: 2413–2423.

  2. Arias AM . (2001). Cell 105: 425–431.

  3. Bogenrieder T, Herlyn M . (2003). Oncogene 22: 6524–6536.

  4. Bowden GT . (2004). Nat Rev Cancer 4: 23–35.

  5. Breitenbach U, Tuckermann JP, Gebhardt C, Richter KH, Furstenberger G, Christofori G et al. (2001). J Invest Dermatol 117: 634–640.

  6. Cavallaro U, Schaffhauser B, Christofori G . (2002). Cancer Lett 176: 123–128.

  7. Chen Y, Miller C, Mosher R, Zhao X, Deeds J, Morrissey M et al. (2003). Cancer Res 63: 1927–1935.

  8. Clarke PA, te Poele R, Wooster R, Workman P . (2001). Biochem Pharmacol 62: 1311–1336.

  9. Cromer A, Carles A, Millon R, Ganguli G, Chalmel F, Lemaire F et al. (2004). Oncogene 23: 2484–2498.

  10. Diaz-Guerra M, Haddow S, Bauluz C, Jorcano JL, Cano A, Balmain A et al. (1992). Cancer Res 52: 680–687.

  11. Diehl F, Grahlmann S, Beier M, Hoheisel JD . (2001). Nucl Acids Res 29: E38.

  12. DiGiovanni J . (1992). Pharmacol Ther 54: 63–128.

  13. Dooley TP, Reddy SP, Wilborn TW, Davis RL . (2003). Biochem Biophys Res Commun 306: 1026–1036.

  14. Egeblad M, Werb Z . (2002). Nat Rev Cancer 2: 161–174.

  15. Furstenberger G, Kopp-Schneider A . (1995). Carcinogenesis 16: 61–69.

  16. Gebhardt C, Breitenbach U, Richter KH, Furstenberger G, Mauch C, Angel P et al. (2005). Am J Pathol 167: 243–253.

  17. Gebhardt C, Breitenbach U, Tuckermann JP, Dittrich BT, Richter KH, Angel P . (2002). Oncogene 21: 4266–4276.

  18. Glinsky GV, Ivanova YA, Glinskii AB . (2003a). Cancer Lett 201: 67–77.

  19. Glinsky GV, Krones-Herzig A, Glinskii AB . (2003b). Neoplasia 5: 218–228.

  20. Gogali A, Charalabopoulos K, Constantopoulos S . (2004). Exp Oncol 26: 106–110.

  21. Hu YC, Lam KY, Law S, Wong J, Srivastava G . (2001a). Clin Cancer Res 7: 2213–2221.

  22. Hu YC, Lam KY, Law S, Wong J, Srivastava G . (2001b). Clin Cancer Res 7: 3519–3525.

  23. Huber W, von Heydebreck A, Sultmann H, Poustka A, Vingron M . (2002). Bioinformatics 18(Suppl. 1): S96–S104.

  24. Ishikawa N, Daigo Y, Yasui W, Inai K, Nishimura H, Tsuchiya E et al. (2004). Clin Cancer Res 10: 8363–8370.

  25. Ito N, Hasegawa R, Imaida K, Hirose M, Asamoto M, Shirai T . (1995). Crit Rev Oncol Hematol 21: 105–133.

  26. Kargul GJ, Dudekula DB, Qian Y, Lim MK, Jaradat SA, Tanaka TS et al. (2001). Nat Genet 28: 17–18.

  27. Kerkela E, Saarialho-Kere U . (2003). Exp Dermatol 12: 109–125.

  28. Kettunen E, Anttila S, Seppanen JK, Karjalainen A, Edgren H, Lindstrom I et al. (2004). Cancer Genet Cytogenet 149: 98–106.

  29. Ko MS, Kitchen JR, Wang X, Threat TA, Wang X, Hasegawa A et al. (2000). Development 127: 1737–1749.

  30. Krieg P, Schnapke R, Furstenberger G, Vogt I, Marks F . (1991). Mol Carcinog 4: 129–137.

  31. Leethanakul C, Patel V, Gillespie J, Shillitoe E, Kellman RM, Ensley JF et al. (2000). Oral Oncol 36: 474–483.

  32. Li J, Johnson JA . (2002). Physiol Genomics 9: 137–144.

  33. Lu J, Liu Z, Xiong M, Wang Q, Wang X, Yang G et al. (2001). Int J Cancer 91: 288–294.

  34. Luo A, Kong J, Hu G, Liew CC, Xiong M, Wang X et al. (2004). Oncogene 23: 1291–1299.

  35. Marks F, Furstenberger G . (1990). Carcinogenesis 11: 2085–2092.

  36. Mendez E, Cheng C, Farwell DG, Ricks S, Agoff SN, Futran ND et al. (2002). Cancer 95: 1482–1494.

  37. Nagata M, Fujita H, Ida H, Hoshina H, Inoue T, Seki Y et al. (2003). Int J Cancer 106: 683–689.

  38. Nair KS, Naidoo R, Chetty R . (2005). J Clin Pathol 58: 343–351.

  39. Patil MA, Chua MS, Pan KH, Lin R, Lih CJ, Cheung ST et al. (2005). Oncogene 24: 3737–3747.

  40. Petroziello J, Yamane A, Westendorf L, Thompson M, McDonagh C, Cerveny C et al. (2004). Oncogene 23: 7734–7745.

  41. Rennecke J, Rehberger PA, Furstenberger G, Johannes FJ, Stohr M, Marks F et al. (1999). Int J Cancer 80: 98–103.

  42. Rundhaug JE, Gimenez-Conti I, Stern MC, Budunova IV, Kiguchi K, Bol DK et al. (1997). Mol Carcinogen 20: 125–136.

  43. Schlingemann J, Hess J, Wrobel G, Breitenbach U, Gebhardt C, Steinlein P et al. (2003). Int J Cancer 104: 699–708.

  44. Stern MC, Duran HA, McKenna EA, Conti CJ . (1997). Mol Carcinogen 20: 137–142.

  45. Strickland JE, Greenhalgh DA, Koceva-Chyla A, Hennings H, Restrepo C, Balaschak M et al. (1988). Cancer Res 48: 165–169.

  46. Tamayo P, Slonim D, Mesirov J, Zhu Q, Kitareewan S, Dmitrovsky E et al. (1999). Proc Natl Acad Sci USA 96: 2907–2912.

  47. Tanaka TS, Jaradat SA, Lim MK, Kargul GJ, Wang X, Grahovac MJ et al. (2000). Proc Natl Acad Sci USA 97: 9127–9132.

  48. Tsai WC, Tsai ST, Ko JY, Jin YT, Li C, Huang W et al. (2004). Oral Oncol 40: 418–426.

  49. Wrobel G, Schlingemann J, Hummerich L, Kramer H, Lichter P, Hahn M . (2003). Nucl Acids Res 31: e67.

  50. Young MR, Nair R, Bucheimer N, Tulsian P, Brown N, Chapp C et al. (2002). Mol Cell Biol 22: 587–598.

  51. Young MR, Yang HS, Colburn NH . (2003). Trends Mol Med 9: 36–41.

  52. Yuspa SH . (1998). J Dermatol Sci 17: 1–7.

Download references


We thank Gunnar Wrobel, Bjoern Tews and Grischa Toedt for statistical and IT support as well as Ingeborg Vogt, Inga Ruiner, Heidi Kramer and Daniela Bodemer for technical assistance. This study was supported by the German Ministry for Education and Research (National Genome Research Network, NGFN-1, 01 GR 0101, NGFN-2, 01 GS 0460) to PL and MH, the Deutsche Forschungsgemeinschaft (AN 182/8-2), by the Research Training Network (RTN) Program of the European Community to PA and the Centre of Molecular Medicine, University of Cologne (BMFT/IDZ 10, Grant 01 GB 950/4) to CM.

Author information



Corresponding author

Correspondence to P Lichter.

Additional information

Supplementary Information accompanies the paper on Oncogene website (

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Hummerich, L., Müller, R., Hess, J. et al. Identification of novel tumour-associated genes differentially expressed in the process of squamous cell cancer development. Oncogene 25, 111–121 (2006).

Download citation


  • multistage carcinogenesis
  • papilloma
  • squamous cell carcinoma
  • gene expression profiling
  • self-organizing maps

Further reading