ORIGINAL ARTICLE
Superficial, Nodular, and Morpheiform Basal-Cell Carcinomas Exhibit Distinct Gene Expression Profiles Mei Yu1,2, David Zloty1, Bryce Cowan1, Jerry Shapiro1, Anne Haegert3, Robert H. Bell3, Larry Warshawski1, Nicholas Carr4 and Kevin J. McElwee1,2 Basal-cell carcinoma (BCC), the most common neoplasm in humans, occurs in a variety of morphological presentations. The mechanisms of BCC development downstream of the initial genetic mutations are not well understood, and different BCC morphological presentations might exhibit distinct gene expression patterns. We investigated superficial (n ¼ 8), nodular (n ¼ 8), and morpheiform (n ¼ 7) BCCs using 21K cDNA microarrays. Global gene expression profiles between respective BCC subtypes, and as compared with normal skin (n ¼ 8), were statistically defined by significance analysis of microarrays (SAM). Thirty-seven genes were subsequently validated by quantitative reverse transcriptase-PCR analysis using an expanded set of 31 BCCs. Gene ontology analysis indicated that gene expression patterns of BCC subtypes in multiple biological processes showed significant variation, particularly in genes associated with the mitogen-activated protein kinase (MAPK) pathway. Notably, genes involved in response to DNA-damage stimulus were uniquely upregulated in morpheiform BCCs. Our results indicate a relative similarity in gene expression between nodular and superficial BCC subtypes. In contrast, morpheiform BCCs are more diverse, with gene expression patterns consistent with their more ‘‘invasive’’ phenotype. These data may help us understand the complex behavior of BCC subtypes and may eventually lead to new therapeutic strategies. Journal of Investigative Dermatology (2008) 128, 1797–1805; doi:10.1038/sj.jid.5701243; published online 17 January 2008
INTRODUCTION Basal-cell carcinoma (BCC) (Pinkus, 1970; Kruger et al., 1999) is the most common neoplasm of humans (Miller, 1991a, b), which arises from the basal layer of the epidermis or the pilosebaceous adnexa (Kruger et al., 1999). Although BCCs are a non-metastatic malignancy, if left untreated, the tumors can invade locally, causing significant destruction to the surrounding soft tissues (Armstrong and Kricker, 2001). Histopathologically, BCCs can be classified into nodular, micronodular, superficial, infiltrative, morpheiform, and mixed BCC forms (Sloane, 1977; Sexton et al., 1990; Rippey, 1998). Clinical and histological subtypes of BCC may exhibit different patterns of behavior and may even have a different etiology (Betti et al., 1995; McCormack et al., 1997; Bastiaens et al., 1998; Scrivener et al., 2002; De Vries
1
Department of Dermatology and Skin Science, University of British Columbia, Vancouver, British Columbia, Canada; 2Vancouver Coastal Health Research Institute, Vancouver, British Columbia, Canada; 3Prostate Centre, Vancouver General Hospital, Vancouver, British Columbia, Canada and 4 Department of Surgery, University of British Columbia, Vancouver, British Columbia, Canada Correspondence: Dr Kevin J. McElwee, Department of Dermatology and Skin Science, University of British Columbia, 835 W. 10th Avenue, Vancouver, British Columbia, Canada V5Z 1L8. E-mail:
[email protected] Abbreviations: BCC, basal-cell carcinoma; COX2, cyclooxygenase-2; GO, gene ontology; Hh, Hedgehog; MAPK, mitogen-activated protein kinase; qPCR, quantitative PCR; SAM, significance analysis of microarrays Received 6 March 2007; revised 23 October 2007; accepted 14 November 2007; published online 17 January 2008
& 2008 The Society for Investigative Dermatology
et al., 2004). The distinction between these different BCC types is important for prognosis and treatment, as more aggressive therapy might be necessary for specific BCC variants (Orengo et al., 1997). Activation of the Hedgehog (Hh) pathway is required in the pathogenesis of most BCCs (Bale and Yu, 2001; McMahon et al., 2003). Somatic alterations of Sonic Hh (SHH), receptor Patched (PTCH1), and SMOH have been identified in BCCs (Gailani et al., 1996; Reifenberger et al., 1998; Lam et al., 1999). In addition, mice ectopically expressing the human Cubitus interruptus homolog GLI-1 in the skin develop tumors closely resembling human BCCs (Nilsson et al., 2000). The common effect of these genetic alterations is constitutive activation of the Hh pathway and transduction of the target genes. Furthermore, 56% of BCCs also present with p53 mutations (Soehnge et al., 1997). While the Hh pathway plays the defining role in the fate of BCC development, the downstream consequences of its constitutive activation are less well defined. How the Hh pathway activation ultimately leads to a BCC phenotype is relatively poorly understood. Although the nature of different cancer phenotypes is diverse, it has been shown that there are relatively few signaling pathways common to cancer formation (Kastan et al., 1991; Hanahan and Weinberg, 2000; Futreal et al., 2004). Modulation of these pathways through differential gene expression elicits the hallmarks of cancer: apoptosis evasion, self-sufficiency in growth signals, angiogenesis, tissue invasion, and unlimited replication (Hanahan and www.jidonline.org 1797
M Yu et al. Basal-Cell Carcinoma Subtypes
Weinberg, 2000). Although contribution of all these hallmarks to BCC development may be debatable given that BCCs are typically non-metastatic, it is likely that at least some gene pathways defined in other forms of cancer will also play a key role in BCC tumorigenesis. Examining the changes in biologically meaningful sets of genes within these pathways may provide a better understanding of the biological process underlying BCC development (Wong et al., 2003). Recent advances in genome technologies have provided an excellent opportunity to determine the complete biological characteristics of neoplastic tissues, potentially resulting in improved diagnosis and development of superior therapeutic strategies. DNA microarray analysis can accomplish this objective and establish sophisticated algorithms (Bertucci et al., 2001; Ramaswamy and Golub, 2002). Using DNA microarrays has also provided important insights into the molecular heterogeneity of cancers and identification of individual genes whose expression is associated with prognosis (Beer et al., 2002; Jenssen et al., 2002; Pomeroy et al., 2002; Rosenwald et al., 2002; Shipp et al., 2002; van’t Veer et al., 2002). Two microarray analyses of BCC have been published. The first study involved a limited cDNA microarray representing 1,718 genes and consequently provided a restricted data set for analysis (Howell et al., 2005). The second, recent study comprehensively evaluated 54,675 gene transcripts in 20 BCC samples versus five normal skin samples (O’Driscoll et al., 2006). However, no published microarray studies have evaluated different subtypes of BCC and compared their respective gene expression profiles. We proposed that the phenotypic diversity of BCCs might be accompanied by a corresponding diversity in gene
Normal
Morpheiform
expression patterns, which could be captured using cDNA microarrays. By evaluating histologically confirmed superficial, nodular, and morpheiform BCCs, we anticipated the identification of genes and signaling pathways common to all three BCC subtypes, which may define the nature of BCC neoplasias. In addition, we anticipated defining gene expression specific for each BCC subtype and possibly elucidating new avenues for future therapeutic treatment development. RESULTS Comparison of sample expression profiles by hierarchical clustering
A hierarchical clustering of unfiltered data from each tissue sample was conducted (Figure 1). Most of the conditions (normal epithelium, superficial BCCs, nodular BCCs, morpheiform BCCs) segregated well in this method. Normal skin epithelium was clearly separated from all BCC subtypes with distinct gene expression profiles, but with close similarity within the group. Superficial and nodular tissue samples also yielded relatively distinctive expression profiles. However, morpheiform BCC expression profiles tended to be distributed among the superficial and nodular tumor types, and were less readily identified by computer analysis as a distinct entity. This is most likely due to a relative heterogeneity of this kind of BCC subtype. Overall, the data support the clinical and histological BCC subtype distinction at the molecular level. Gene expression common to all BCC subtypes versus skin epithelium
Gene expression across all BCC subtypes together (all 23 BCC samples as a single group), as compared with normal
Nodular
Superficial
Figure 1. Hierarchical clustering of samples. Unfiltered raw microarray data from each sample of normal skin epithelium (circles), superficial BCC (stars), nodular BCC (triangles), and morpheiform BCC (squares) were clustered by Pearson correlation and distances between clusters were calculated by average linkage. Normal epithelium samples were distinct from BCCs and closely similar in expression profiles. Of BCC subtypes, morpheiform BCCs exhibited the greatest heterogeneity in gene expression.
1798 Journal of Investigative Dermatology (2008), Volume 128
M Yu et al. Basal-Cell Carcinoma Subtypes
skin epithelium, was evaluated by the significance analysis of microarray (SAM) method with a cut-off q-value of 36.84% equating to a 95% confidence level. The results indicated that 889 genes were significantly differentially expressed in BCCs. Of these, 88 were upregulated and 801 downregulated as compared with normal skin epithelium. Table S2a and b display the top 25 most differentially expressed probe sets in all three BCC subtypes combined, compared with normal skin epithelium. Differential gene expression in BCC subtypes
The most differentially expressed genes compared with normal skin epithelium were then identified within each of the three different groups of BCC subtypes: superficial, nodular, and morpheiform. After analysis of data by the SAM method, we identified 4,037 probe sets (set 1) as being differentially expressed between nodular BCCs and normal skin epithelium, 1,914 probe sets (set 2) between morpheiform BCCs and normal skin epithelium, and 397 probe sets (set 3) between superficial BCCs and normal skin epithelium. Table S3a–f display the 20 most differentially expressed probe sets ordered by decreasing evidence for differential gene expression between each BCC subtype, compared with normal skin epithelium. Differential gene expression common to all BCC subtypes versus normal skin epithelium
Based on comparison between gene sets 1, 2, and 3, we obtained a common set of genes (set 4) expressed in all three BCC subtypes. In all, 164 genes were identified by the SAM method in all three BCC subtypes, as compared with normal skin epithelium. The top 20 common genes identified with significant differential expression in all three BCC subtypes, compared with normal skin epithelium, are listed (Table S4). Unique and differential BCC subtype gene expression analyses
Cross-comparison of gene lists (analysis of differentially and similarly expressed gene sets derived from sets 1, 2, and 3; see Materials and Methods) for each BCC subtype was conducted to identify those genes with statistically significant expression unique to only one of the three BCC subtypes. We identified 11, 343, and 94 genes with a statistically significant fold change in expression unique to superficial, nodular, or morpheiform BCCs, respectively (Table S5a–f). In principle, these tables of differential gene expression may represent signal transduction pathways and mechanisms which might determine the specific phenotype of each respective BCC subtype. In addition, we conducted pairwise comparisons between conditions using the SAM method to define differential gene expression between BCC subtypes, without relation to gene expression in normal skin epithelium, and between all BCCs and skin epithelium. For each comparison, we report the 25 up- and downregulated genes that had the lowest expected false discovery rate (Table S6a–h). Ontological analysis of BCC subtypes
To analyze whether genes with specific functions were over- or under-expressed in the three subtypes of BCCs, we
analyzed each subtype’s composition of differentially expressed genes separately (gene sets 1, 2, and 3) with respect to their gene ontology (GO annotation), as given in the DAVID annotation system. Of significant interest, some GO categories were uniquely represented in respective BCC subtypes (Table S7). Only one GO class, ‘‘external stimuli function’’, was defined as being uniquely expressed in superficial BCC subtypes but not nodular or morpheiform BCCs. Gene enrichment in the categories of ‘‘regulation of cell growth’’ and ‘‘negative regulation of enzyme activity’’ were characterized in nodular BCCs, among others. In contrast, two functional GO categories, ‘‘response to DNA damage stimulus’’ and ‘‘ectoderm development’’, were uniquely identified with statistically significant expression in morpheiform BCCs, but not superficial or nodular BCCs. Tumor progression pathway analysis
Recent publications have promoted the value of examining gene expression data in terms of signal transduction networks represented rather than individual genes expressed (Wong and Chang, 2005). Using gene sets 1, 2 and 3, classified respectively in terms of their GO annotated roles, we defined the statistical significance of pathways and categories of genes represented using Fisher’s gene-enrichment calculation and subsequently further defined more specific descriptions (GO level 5). By subdividing the list into distinct sets of genes that were pathway-specific, we found that the most strongly represented bias, in terms of the number of genes differentially expressed across all three BCC subtypes, was toward apoptosis and its regulation (Table S8). Focusing only on genes involved in established pathways of tumorigenesis (the Hh, Wnt, transforming growth factor-b, apoptosis, calcium ion channel, mitogen-activated protein kinase (MAPK), or cell-cycle-signaling pathways), we identified 93 genes (using a 36.84% SAM q-value cut-off) that were differentially expressed between all BCCs and normal skin, and were also involved in one or more of the seven identified pathways. Seventeen genes were involved in the Hhsignaling pathway, 27 in the Wnt-signaling pathway, 22 in the transforming growth factor-b-signaling pathway, 89 in apoptosis, 18 in calcium channel signaling, 22 in MAPK, and 21 in cell cycling. Upon comparing every two BCC subtypes with each other, we found that the predominant signal transduction pathway with the greatest difference in the number of differentially expressed, pathway-associated genes was the MAPK pathway (data not shown). The MAPK pathway is of significant interest in understanding BCC subtypes, given its interaction with many cellular systems and other signal transduction pathways, including Hh and Wnt pathways. Validation of the expression of selected genes
With the microarray analysis of the three BCC subtypes complete, we selected genes with known functional significance, primarily within the apoptosis and MAPK pathway categories, for evaluation by quantitative PCR (qPCR). Selected genes, particularly those associated with the Hh- and cell-cycle-signaling pathways, were also evaluated, www.jidonline.org 1799
M Yu et al. Basal-Cell Carcinoma Subtypes
35
Nodular
30
*
Fold change
20
All BCCs
*
c
* *
15
b
*
Morpheiform
25
a
*
Superficial
10 5 1 COX2
AIF1
TNRFSF10D
Figure 2. Characterization of gene expression patterns across BCC subtypes. COX2, AIF1, and TNRSF10D were selected as examples of genes defined by microarray as having increased fold change expression in the morpheiform, nodular, or superficial BCC subtypes, respectively. Plotting of qPCR validation results confirmed that COX2 was statistically significantly expressed in morpheiform BCCs, whereas AIF1 was most highly expressed in nodular BCCs and TNRFSF10D was highly expressed in superficial BCCs (* ¼ Po0.05). AIF1, apoptosis-inducing factor 1.
even though they did reach statistical significance in the SAM analysis. Although the fold change in expression defined by qPCR was different from that observed by microarray, the trends, whether for increased or decreased gene expression, were mostly consistent for each respective gene examined and the majority of evaluated genes were expressed with statistically significant differences in expression (Table S9). The majority of genes presented with a modest fold change in expression in superficial BCCs, but with greater differential expression in nodular and morpheiform BCCs. A similar trend pattern emerged for many genes evaluated by qPCR. However, a few genes identified with the greatest fold change in gene expression in nodular and superficial BCC subtypes by microarray were confirmed by qPCR as most highly expressed in the respective BCC subtypes (Figure 2). Variability in mRNA expression within BCC subgroups was apparent, but results were statistically significant between groups. Such differences in gene expression might potentially be utilized as molecular identifiers of specific BCC subtypes. As a validation of product expression signaled by microarray and qPCR results, we conducted immunohistology to define cyclooxygenase-2 (COX2) expression in pathology specimens from the superficial, nodular, and morpheiform BCCs. The results were consistent with the pattern of COX2 mRNA expression observed. Low, or almost no, COX2 was consistently seen in superficial BCCs (Figure 3a). Expression in nodular BCCs varied with location and specimen, but overall suggested a moderate presence (Figure 3b), whereas COX2 was readily identified in the invasive threads of morpheiform BCCs (Figure 3c). DISCUSSION Comparison to previously published microarray analyses
Thus far, no published studies have compared the gene expression profiles of different, histologically confirmed BCC 1800 Journal of Investigative Dermatology (2008), Volume 128
Figure 3. COX2 expression in BCC subtypes. Immunohistochemistry consistently identified almost no expression of COX2 in superficial BCCs (a), and moderate expression in nodular BCCs (b), whereas tongues of neoplastic basaloid cells in morpheiform BCCs in relatively collagenized stroma exhibited consistent COX2 expression (c). Bars ¼ 50 mm.
subtypes using microarrays. Only two microarray-based studies have been published, which compare BCC tissues with normal skin epithelium (Howell et al., 2005; O’Driscoll et al., 2006). Comparison of our data with already published datasets identify a number of confirmatory observations when considering that our gene expression sets observed are common to all three BCC subtypes (Table S2a and b). Notably, some genes are well represented both in previous studies and in our current evaluation, which include the following: collagen, type IV, a1 (COL4A1), collagen, type V, a2 (COL5A2), collagen, type VI, a1 (COL6A1), syndecan 2 (SDC2), and tumor-associated calcium signal transducer 1 (TACSTD1), among others. However, despite the notable confirmation of a few highly expressed genes between the two studies (such as chromogranin A (CHGA)), comparison of our datasets with those of O’Driscoll and co-workers suggests a greater number of differences between the gene expression profiles. How much of this difference is due to the differences in the cDNA microarray platforms and statistical analysis approaches used is not known. Most likely, the differences in the tissue samples and the nature of BCC subtypes utilized in each study have determined the distinctions in results. Gene expression and function across BCC subtypes
To analyze whether genes with specific functions were overor under-expressed in the three subtypes of BCCs, we analyzed each subtype’s composition of the filtered and tested genes with respect to their GO annotation. Several GO classifications were well represented in all three BCC subtypes, including cell motility, cellular morphogenesis, negative regulation of cellular process, positive regulation of cellular process, and cellular metabolism. These classifications have all been suggested to play significant roles in carcinogenesis in different tumors (Kunz-Schughart et al., 2000; Mazars et al., 2000; Jeon et al., 2004; Liou et al., 2004; Zhivotovsky and Orrenius, 2006). However, within a category, each BCC subtype had different numbers of genes and different degrees of fold change in expression. By both
M Yu et al. Basal-Cell Carcinoma Subtypes
microarray technique and qPCR, most genes evaluated were found with moderate increased expression in superficial BCCs, but with a much greater level of differential expression in nodular and morpheiform BCCs. Examples of this trend included MMP1, DPP4, and ATP2B1 (Table S9). While genes may be commonly expressed in all three BCC subtypes, their different levels of expression will be important for determining differences in BCC subtype morphology. The gene function classes enriched in superficial BCC subtypes showed a significant association with ‘‘external stimuli function’’ and possibly ‘‘positive regulation of enzyme activity’’ (Table S7). In nodular BCCs, gene enrichment in the categories ‘‘regulation of cell growth’’, ‘‘cell growth’’, and ‘‘chromosome segregation’’ were identified, among others. The combination of these functional categories might be relevant for development of the nodular phenotype. Two functional GO categories, ‘‘ectoderm development’’ and ‘‘response to DNA damage stimulus’’, were uniquely identified with statistically significant expression exclusively in morpheiform BCCs. Genes from these categories such as inducible enzyme COX2 might be a focus of interest. Studies have suggested that overexpression of COX2 might induce the expression of vascular endothelial growth factor, increase angiogenesis, and enhance tumor growth, metastasis, and tumor spread (Tong et al., 2000; Uefuji et al., 2000; Ferrandina et al., 2002; Li et al., 2006). We found that COX2 was somewhat expressed in nodular BCCs, but much more so in morpheiform BCCs, by qPCR. We were also able to confirm a difference in the presence of COX2 product in the three BCC subtypes, by immunohistology. The expression of genes associated with angiogensis in morpheiform BCCs may reflect the more aggressive/invasive activity of this BCC subtype, and could be markers for diagnosis and perhaps a target for future tailor-made therapeutic strategies. Hh and Wnt signal transduction pathways
The underlying signal transduction pathways known to be functionally important in BCC growth, the Hh and Wnt signal transduction pathways (McMahon, 2000), were detected in all three subtypes of BCCs. With the notable exception of Gli2, the stringency of the microarray analysis resulted in a lack of statistical significance in many Hh pathway-associated genes, although greater than 1.5-fold mean changes in expression were apparent. However, several Hh pathway-associated genes were identified with significant differential expression in all BCC subtypes. GLI transcription regulators AKT1, IGFBP4, IGFBP2, and PRKACA were upregulated in all BCCs (Epstein et al., 1996; Hammerschmidt et al., 1996; Riobo et al., 2006; Sheng et al., 2006). Subsequent evaluation by the more sensitive qPCR technique confirmed upregulation of PTCH1, GLI1, GLI2, and PRKACA in BCC subtypes (Table S9). Other Hh pathway-associated genes identified in microarray results and re-evaluated by qPCR included SFRP2 and BMPR2, both of which exhibited upregulation in the BCCs, with fold change greater in nodular and morpheiform types (Table S9). Expression of Wnt family members, including CCND2, CAMK2G, CSNK2A2, and especially frizzled receptors for
Wnt proteins, FZD7, FZD8, and FZD2, was deregulated in BCCs. These genes are known to play a role in several types of cancers (Janssens et al., 2004; Milovanovic et al., 2004; Vincan et al., 2005). By microarray analysis, levels of differential expression for most Wnt pathway-associated genes were found to be similar across the three BCC subtypes. However, by qPCR, WNT5A and FZD2 were observed with the greatest fold increase in nodular BCCs and the least in superficial BCCs (Table S9). Overall, these changes in expression are consistent with the involvement of the canonical Wnt-signaling pathway in BCCs. In addition, we found evidence consistent with the activation of the noncanonical Wnt/Ca2 þ pathway (Miller et al., 1999; Kuhl et al., 2000; Ishitani et al., 2003). Apoptosis pathway
There are two major apoptotic pathways initiated by either mitochondria (‘‘intrinsic’’ pathway) or death receptors (‘‘extrinsic’’ pathway) (Engels et al., 2000; Sorlie et al., 2001; Liu et al., 2004). Our microarray data showed that genes in both pathways were differentially expressed, although statistical significance was not always achieved using SAM analysis. For example, differential fold change expression of decoy TRAIL receptor 4 (TNFRSF10D), which protects against TRAIL-mediated apoptosis, failed to reach statistical significance in SAM analysis. However, by qPCR there was statistically significant upregulation, most notably with highest expression in superficial BCCs. TNFRSF6B, coding for DcR3, which binds to FasL inhibiting its proapoptotic action (Tsuji et al., 2003), was upregulated much more so in nodular and morpheiform BCCs than in superficial BCCs (Table S9). Many genes have been implicated in the process of mitochondria-mediated apoptosis, including the Bcl family, the inactivation of which might contribute to tumor progression (Huang, 2000; Antonsson, 2001). Consistent with Bcl family-associated mechanism activity, we found reduced levels of BCL2L2, BCL3, BCL6, and BCL7B. In contrast, BCL2L11, BCL7C, BCL10, and BCL11B were upregluated along with somewhat increased levels of upstream proto-oncogene AKT1. Overall, the results suggested that BCC cells may exhibit mechanisms to evade apoptosis, particularly in morpheiform BCCs, allowing cell survival and increased proliferation (Erb et al., 2005). MAPK-signaling pathway
MAPKs are evolutionary conserved enzymes connecting cellsurface receptors to critical regulatory targets within cells (Strniskova et al., 2002). Mammals express at least four distinctly regulated MAPK components, extracellular signalrelated kinases, Jun amino-terminal kinases (JNK1/2/3), p38 proteins (p38a/b/g/d) and extracellular signal-related kinase-5 (Chang and Karin, 2001). Of the members of the MAPKsignaling pathway, Ras and Raf have been identified as protooncogenes, where gain-of-function (activating) mutations have been identified in many cancers (Hussein, 2005). We found increases in the expression of many Ras superfamily members in all three BCC subtypes, such as RAB31, which has also been identified in breast cancer (Abba et al., 2005), www.jidonline.org 1801
M Yu et al. Basal-Cell Carcinoma Subtypes
and RAB3A, which is enhanced in harderian tumors in mice (Vadlamudi et al., 2000). In contrast, tumor suppressor genes such as DUSP2 and RASSF1 (Akino et al., 2005) were reduced in expression. These results and others suggested that the RAS-MAPK-signaling pathway is significantly activated and may play a key role in BCC carcinogenesis. However, we were unable to obtain statistically significant results for RAB3A and DUSP2 by qPCR. Nevertheless, given that receptors associated with Ras activation seem to be effective targets for treating breast cancer (Slamon et al., 2001), this family may be worthy of further investigation. Conclusions
This study identifies genome-wide gene expression profiles of superficial, nodular, and morpheiform BCC subtypes. Our results indicated that several of the signaling pathways identified in other forms of cancer are well represented in BCCs. Some specific categories of genes were identified in specific BCC subtypes and these categories/genes may determine the particular clinical and histological BCC phenotypes. Where genes were commonly expressed in all three BCC subtypes, they typically presented with greatest fold change in expression in morpheiform BCCs, suggesting that morpheiform BCCs are more ‘‘aggressive’’ compared with superficial and nodular BCCs. Many of the genes defined may regulate pathways essential for BCC formation in human skin and could eventually represent potential therapeutic targets. Further exploration of the relevant mechanisms of pathogenesis will improve our understanding of the molecular basis of BCC. MATERIALS AND METHODS BCC tissues and clinical information Human BCCs and normal skin tissues were obtained from 31 patients at the Department of Dermatology and Skin Science, University of British Columbia. Investigations were performed according to the Declaration of Helsinki Principles after approval by the University Clinical Research Ethics Committee. Specific patient consent was not required because Canadian law considers human tissue excised during surgery as discarded material. For microarray analysis, superficial (8), nodular (8), morpheiform (7), and microdissected normal skin epithelium (8) tissue samples were collected. All of the samples were taken from the facial area. Of these 31 samples, 18 were derived from males (45–75 years of age with a mean±SD of 57.4±7.90 years) and 13 from females (43–84 years of age with a mean±SD of 59.4±11.73 years). Samples were collected in the operation theater during Moh’s surgical excision and were immediately stored in an RNA stabilization reagent (Qiagen Inc., Mississauga, ON, Canada). Only tissues from patients who had not been treated with preoperative chemotherapy or other therapeutic approaches were selected. BCC morphological subtypes were described and classified during surgery, and clinical diagnoses were subsequently confirmed by formalin-fixed, paraffin-embedded histological assessment of the tumors.
RNA isolation Total RNA was isolated with an RNeasy Fibrous Tissue Midi kit (Qiagen) according to the manufacturer’s protocols. The quantity 1802 Journal of Investigative Dermatology (2008), Volume 128
and quality of the RNA were measured using an Agilent 2100 bioanalyzer and the RNA 6000 NANO kit (Agilent Technologies, Palo Alto, CA).
Microarray production Human Operon v.2.1 (21K) glass arrays were designed (based on human 70-mer from Operon Biotechnologies Inc., Huntsville, AL) by the Microarray Facility of the Prostate Centre at Vancouver General Hospital, Vancouver, Canada (Lyons, 2003; Nelson et al., 2003). RNAs were amplified using the SenseAmp Plus kit (Genisphere Inc., Hatfield, PA). The calculated A 260/280 ratio was used to determine the appropriate amount of sense RNA for labeling. Total RNA from test samples and universal human reference RNA (Stratagene, Cedar Creek, TX) were differentially labeled with Cy5 and Cy3, respectively, using the 3DNA array detection 350 kit (Genisphere Inc.), and co-hybridized to cDNA microarrays. Following overnight hybridization and washing, arrays were imaged using a ScanArray Express scanner (PerkinElmer, Boston, MA).
Data processing and analysis Arrays were scanned at excitation wavelengths of 532 and 635 nm to detect the Cy3 and Cy5 dyes, respectively. Image analysis and quantification were conducted with commercial software (Imagene 6.0 software; Biodiscovery Inc., El Segundo, CA). After grid assignment, the adjusted intensity for each gene was calculated by subtracting the background median from the signal median. This value was used as input for the Genespring 7.2 program (Silicon Genetics, Redwood City, CA), which allows multiple filter comparisons using data from different experiments to perform normalization, generation of restriction lists, and the functional classification of the differentially expressed genes. Raw data were used in ‘‘per chip and per spot normalization’’, which was intensity-dependent (non-linear or LOWESS normalization) (Yang et al., 2002). The expression of each gene is reported as the ratio of the value obtained after each condition relative to control conditions after normalization of the data. Data were subsequently filtered using the raw signal strength value. Measurements with higher signal strength value are relatively more precise than measurements with lower control strength. Consequently, genes with expression that did not reach the minimum signal value (100 arbitrary units) were discarded. We generated a condition tree using hierarchical clustering of unfiltered data from each sample, based on the similarity of their expression data. Similarity was measured using Pearson correlation and distances between clusters were calculated by average linkage. All raw data from the arrays have been entered into the public Gene Expression Omnibus (GEO) database in MIAME compliant format (http://www.ncbi.nlm.nih.gov/geo/). The raw data sets are included under the series record number GSE6520.
Analysis of gene expression differences between BCC subgroups and unique to BCC subtypes SAM was used to identify genes differentially regulated between skin epithelium and the respective subtypes of BCCs. SAM is a statistical technique for finding significant genes in a set of microarray experiments (Tusher et al., 2001). It assigns a score to each gene on the basis of change in the gene expression relative to the standard deviation of repeated measurement. It then uses permutations of the repeated measurements to estimate the false discovery rate. After
M Yu et al. Basal-Cell Carcinoma Subtypes
filtering, three respectively statistically significant difference gene lists (sets 1, 2, and 3, see below) between BCC subtypes and skin epithelium were identified. To find genes common to all three subtypes (Table S4), they had to pass a false discovery-corrected q-value of 36.84% for every subtype compared with normal skin individually. This ensured that the chance that a gene made it into the list without having any relevance to any of the subtypes was below 5% (95% confidence interval). This was calculated from the initial 36.84% by 0.36843 (cubed) ¼ 0.05. To define the degree of similarity in gene lists derived from different subtypes of BCCs compared with skin epithelium, the q-value for one subtype had to be below 36.84% and above 100–36.84% ¼ 63.16% in the other subtypes (Table S5a–f). For comparison of gene expression in BCC subtypes, we also used the SAM technique to identify the gene lists (Table S6).
using a total of 25 ml of reaction mixture (2 ml of cDNA, 12.5 ml of 2 SYBR Green PCR Master Mix (Qiagen), 1.5 ml of each 5 mmol l1 forward and reverse primers (Invitrogen), and 7.5 ml of H2O) in an Opticon DNA Engine (MJ Research, Waltham, MA). The PCR program was initiated for 10 minutes at 95 1C before 40 thermal cycles, each of 15 seconds at 94 1C, 30 seconds at 60 1C, and 30 seconds at 72 1C. Preliminary research comparing housekeeping genes b-actin, glyceraldehyde-3-phosphate dehydrogenase, and QuantumRNA Universal 18S internal standard primers (Ambion Inc., Austin, TX) in collected tissues by reverse transcriptase-PCR, identified 18S as most consistently expressed in skin (data not shown). 18S was subsequently used for all analyses described. Reverse transcriptase-PCR data were analyzed according to the comparative Ct method, and were normalized against 18S expression in each sample. Melting curves for each PCR reaction were generated to ensure the purity of the amplification product.
GO and pathway analysis The functional classification of genes is based on the GO database and allows the identification of ‘‘enriched’’ or ‘‘depleted’’ gene function categories in assigned biological processes, molecular functions, and cellular components (Harris et al., 2004). We analyzed the GO annotation for the genes that were significantly different between the BCCs and normal skin, using a web-based annotation tool (DAVID version 2 software, http://david.abcc. ncifcrf.gov/) for identified ‘‘enriched’’ function annotations. This program identifies genes belonging to different GO categories and also calculates the statistical significance of non-random representation (that is, overlapping P-values) with Fisher’s exact test. A P-valuep0.01 indicates that the user gene list is specifically associated (enriched) in this category, as compared with random chance. Potential false positives in our gene list, which can result from technical insufficiencies and low case numbers, become randomly distributed among the GO classes and thus any significant association identified should result from the distribution patterns of the true positives. For 25–40% of the genes defined, no GO annotation was given and their function is unknown. The numbers of genes identified in major categories were normalized by the number of genes annotated in each list and were expressed as percentages (Dennis et al., 2003). To assess which signaling pathways were affected during gene regulation, the genes were classified and grouped into different pathways using Genespring software. Pathway information was imported from the Kyoto encyclopedia of Genes and Genomes.
qPCR analysis for gene expression To verify the alterations of gene expression at the mRNA level identified by microarray, we chose representative genes with varying expression profiles for qPCR. The sequences of the complete genome were obtained from GenBank. A pair of oligonucleotide primers was designed for each sequence using Primer 3 software (http://frodo. wi.mit.edu/cgi-bin/primer3/primer3_www.cgi), using the following criteria: 50–65 1C melting temperature, 40–60% G þ C content, 18–25-bp primer length, and 75–250-bp amplicon size. Primer sequences are listed in Table S1. Two micrograms of total RNA from each sample were subjected to reverse transcription using the Superscript first-strand cDNA synthesis kit (Invitrogen Life Technologies, Carlsbad, CA) according to the manufacturer’s protocol. qPCR reactions were then conducted
Immunohistology Protein expression COX2 was explored in BCC subtypes by immunohistochemistry. Briefly, paraffin-embedded tissue sections were processed, rehydrated, and antigen-unmasked by immersion in 10 mM sodium citrate buffer at 90 1C. After blocking with normal rabbit serum and alkaline phosphatase ablation with levamisole (Vector Laboratories, Burlingame, CA), rabbit anti-human COX2 (Cell Signaling Technology, Danvers, MA), dextran polymerconjugated anti-rabbit, alkaline phosphatase amplification polymer, and substrate (all from Vector Laboratories) were applied in sequence. Sections were counterstained with Harris’ hematoxylin and mounted in Permount (Fisher Scientific, Ottawa, ON, Canada). CONFLICT OF INTEREST The authors state no conflict of interest.
ACKNOWLEDGMENTS This work was financially supported by the Canadian Dermatology Foundation. We thank the staff of the Microarray Facility of The Prostate Centre at Vancouver General Hospital for supplying us with microarrays and for their technical assistance and advice used in this study. We also thank Drs H Lui and S Le Bihan for help and advice.
SUPPLEMENTARY MATERIAL Table S1. Primer sequences for defined genes. Table S2a. Gene transcripts with significant upregulation common to BCCs versus normal skin epithelium. Table S2b. Gene transcripts with significant downregulation common to BCCs versus normal skin epithelium. Table S3a. Gene transcripts with significant upregulation in superficial BCCs versus normal skin epithelium. Table S3b. Gene transcripts with significant upregulation in nodular BCCs versus normal skin epithelium. Table S3c. Gene transcripts with significant upregulation in morpheiform BCCs versus normal skin epithelium. Table S3d. Gene transcripts with significant downregulation in superficial BCCs versus normal skin epithelium. Table S3e. Gene transcripts with significant downregulation in nodular BCCs versus normal skin epithelium. Table S3f. Gene transcripts with significant downregulation in morpheiform BCCs versus normal skin epithelium. Table S4a. Common genes identified with significant increased expression in all three BCC subtypes compared with normal skin epithelium.
www.jidonline.org 1803
M Yu et al. Basal-Cell Carcinoma Subtypes
Table S4b. Common genes identified with significant decreased expression in all three BCC subtypes compared with normal skin epithelium. Table S5a. Genes with increased expression exclusively identified in superficial BCCs. Table S5b. Genes with increased expression exclusively identified in nodular BCCs. Table S5c. Genes with increased expression exclusively identified in morpheiform BCCs. Table S5d. Genes with decreased expression exclusively identified in superficial BCCs. Table S5e. Genes with decreased expression exclusively identified in nodular BCCs. Table S5f. Genes with decreased expression exclusively identified in morpheiform BCCs. Table S6a. Gene transcripts with significant upregulation in morpheiform BCCs versus nodular BCCs. Table S6b. Gene transcripts with significant downregulation in morpheiform BCCs versus nodular BCCs. Table S6c. Gene transcripts with significant upregulation in morpheiform BCCs versus superficial BCCs. Table S6d. Gene transcripts with significant downregulation in morpheiform BCCs versus superficial BCCs. Table S6e. Gene transcripts with significant upregulation in nodular BCCs versus superficial BCCs. Table S6f. Gene transcripts with significant downregulation in nodular BCCs versus superficial BCCs. Table S6g. Gene transcripts with significant upregulation in all BCCs versus skin epithelium. Table S6h. Gene transcripts with significant downregulation in all BCCs versus skin epithelium. Table S7. GO analysis of biological processes represented by differential gene expression observed by BCC subtype.
De Vries E, Louwman M, Bastiaens M, de Gruijl F, Coebergh JW (2004) Rapid and continuous increases in incidence rates of basal cell carcinoma in the southeast Netherlands since 1973. J Invest Dermatol 123:634–8 Dennis G Jr, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC et al. (2003) DAVID: database for annotation, visualization, and integrated discovery. Genome Biol 4:P3 Engels IH, Stepczynska A, Stroh C, Lauber K, Berg C, Schwenzer R et al. (2000) Caspase-8/FLICE functions as an executioner caspase in anticancer drug-induced apoptosis. Oncogene 19:4563–73 Epstein DJ, Marti E, Scott MP, McMahon AP (1996) Antagonizing cAMPdependent protein kinase A in the dorsal CNS activates a conserved Sonic hedgehog signaling pathway. Development 122:2885–94 Erb P, Ji J, Wernli M, Kump E, Glaser A, Buchner SA (2005) Role of apoptosis in basal cell and squamous cell carcinoma formation. Immunol Lett 100:68–72 Ferrandina G, Legge F, Ranelletti FO, Zannoni GF, Maggiano N, Evangelisti A et al. (2002) Cyclooxygenase-2 expression in endometrial carcinoma: correlation with clinicopathologic parameters and clinical outcome. Cancer 95:801–7 Futreal PA, Coin L, Marshall M, Down T, Hubbard T, Wooster R et al. (2004) A census of human cancer genes. Nat Rev Cancer 4:177–83 Gailani MR, Stahle-Backdahl M, Leffell DJ, Glynn M, Zaphiropoulos PG, Pressman C et al. (1996) The role of the human homologue of Drosophila patched in sporadic basal cell carcinomas. Nat Genet 14:78–81 Hammerschmidt M, Pelegri F, Mullins MC, Kane DA, Brand M, van Eeden FJ et al. (1996) Mutations affecting morphogenesis during gastrulation and tail formation in the zebrafish, Danio rerio. Development 123:143–51 Hanahan D, Weinberg RA (2000) The hallmarks of cancer. Cell 100:57–70 Harris MA, Clark J, Ireland A, Lomax J, Ashburner M, Foulger R et al. (2004) The gene ontology (GO) database and informatics resource. Nucleic Acids Res 32:D258–61
Table S8. Gene ontology analysis of biological processes represented by differential gene expression observed common to all BCC subtypes.
Howell BG, Solish N, Lu C, Watanabe H, Mamelak AJ, Freed I et al. (2005) Microarray profiles of human basal cell carcinoma: insights into tumor growth and behavior. J Dermatol Sci 39:39–51
Table S9. Selected gene validation results by qPCR with corresponding microarry results for comparison.
Huang Z (2000) Bcl-2 family proteins as targets for anticancer drug design. Oncogene 19:6627–31
REFERENCES
Hussein MR (2005) Ultraviolet radiation and skin cancer: molecular mechanisms. J Cutan Pathol 32:191–205
Abba MC, Hu Y, Sun H, Drake JA, Gaddis S, Baggerly K et al. (2005) Gene expression signature of estrogen receptor alpha status in breast cancer. BMC Genomics 6:37
Ishitani T, Kishida S, Hyodo-Miura J, Ueno N, Yasuda J, Waterman M et al. (2003) The TAK1–NLK mitogen-activated protein kinase cascade functions in the Wnt-5a/Ca(2+) pathway to antagonize Wnt/beta-catenin signaling. Mol Cell Biol 23:131–9
Akino K, Toyota M, Suzuki H, Mita H, Sasaki Y, Ohe-Toyota M et al. (2005) The Ras effector RASSF2 is a novel tumor-suppressor gene in human colorectal cancer. Gastroenterology 129:156–69
Janssens N, Andries L, Janicot M, Perera T, Bakker A (2004) Alteration of frizzled expression in renal cell carcinoma. Tumour Biol 25:161–71
Antonsson B (2001) Bax and other proapoptotic Bcl-2 family ‘‘killer-proteins’’ and their victim the mitochondrion. Cell Tissue Res 306:347–61 Armstrong BK, Kricker A (2001) The epidemiology of UV induced skin cancer. J Photochem Photobiol B 63:8–18 Bale AE, Yu KP (2001) The hedgehog pathway and basal cell carcinomas. Hum Mol Genet 10:757–62 Bastiaens MT, Hoefnagel JJ, Bruijn JA, Westendorp RG, Vermeer BJ, Bouwes Bavinck JN (1998) Differences in age, site distribution, and sex between nodular and superficial basal cell carcinoma indicate different types of tumors. J Invest Dermatol 110:880–4 Beer DG, Kardia SL, Huang CC, Giordano TJ, Levin AM, Misek DE et al. (2002) Gene-expression profiles predict survival of patients with lung adenocarcinoma. Nat Med 8:816–24 Bertucci F, Houlgatte R, Nguyen C, Viens P, Jordan BR, Birnbaum D (2001) Gene expression profiling of cancer by use of DNA arrays: how far from the clinic? Lancet Oncol 2:674–82 Betti R, Inselvini E, Carducci M, Crosti C (1995) Age and site prevalence of histologic subtypes of basal cell carcinomas. Int J Dermatol 34:174–6 Chang L, Karin M (2001) Mammalian MAP kinase signalling cascades. Nature 410:37–40
1804 Journal of Investigative Dermatology (2008), Volume 128
Jenssen TK, Kuo WP, Stokke T, Hovig E (2002) Associations between gene expressions in breast cancer and patient survival. Hum Genet 111:411–20 Jeon GA, Lee JS, Patel V, Gutkind JS, Thorgeirsson SS, Kim EC et al. (2004) Global gene expression profiles of human head and neck squamous carcinoma cell lines. Int J Cancer 112:249–58 Kastan MB, Onyekwere O, Sidransky D, Vogelstein B, Craig RW (1991) Participation of p53 protein in the cellular response to DNA damage. Cancer Res 51:6304–11 Kruger K, Blume-Peytavi U, Orfanos CE (1999) Basal cell carcinoma possibly originates from the outer root sheath and/or the bulge region of the vellus hair follicle. Arch Dermatol Res 291:253–9 Kuhl M, Sheldahl LC, Park M, Miller JR, Moon RT (2000) The Wnt/Ca2+ pathway: a new vertebrate Wnt signaling pathway takes shape. Trends Genet 16:279–83 Kunz-Schughart LA, Doetsch J, Mueller-Klieser W, Groebe K (2000) Proliferative activity and tumorigenic conversion: impact on cellular metabolism in 3D culture. Am J Physiol Cell Physiol 278:C765–80 Lam CW, Xie J, To KF, Ng HK, Lee KC, Yuen NW et al. (1999) A frequent activated smoothened mutation in sporadic basal cell carcinomas. Oncogene 18:833–6
M Yu et al. Basal-Cell Carcinoma Subtypes
Li W, Xu RJ, Zhang HH, Jiang LH (2006) Overexpression of cyclooxygenase-2 correlates with tumor angiogenesis in endometrial carcinoma. Int J Gynecol Cancer 16:1673–8 Liou LS, Shi T, Duan ZH, Sadhukhan P, Der SD, Novick AA et al. (2004) Microarray gene expression profiling and analysis in renal cell carcinoma. BMC Urol 4:9 Liu D, Kang JS, Derynck R (2004) TGF-beta-activated Smad3 represses MEF2dependent transcription in myogenic differentiation. EMBO J 23: 1557–66
chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med 346: 1937–47 Scrivener Y, Grosshans E, Cribier B (2002) Variations of basal cell carcinomas according to gender, age, location and histopathological subtype. Br J Dermatol 147:41–7 Sexton M, Jones DB, Maloney ME (1990) Histologic pattern analysis of basal cell carcinoma. Study of a series of 1039 consecutive neoplasms. J Am Acad Dermatol 23:1118–26
Lyons P (2003) Advances in spotted microarray resources for expression profiling. Brief Funct Genomic Proteomic 2:21–30
Sheng T, Chi S, Zhang X, Xie J (2006) Regulation of Gli1 localization by the cAMP/protein kinase A signaling axis through a site near the nuclear localization signal. J Biol Chem 281:9–12
Mazars A, Tournigand C, Mollat P, Prunier C, Ferrand N, Bourgeade MF et al. (2000) Differential roles of JNK and Smad2 signaling pathways in the inhibition of c-Myc-induced cell death by TGF-beta. Oncogene 19: 1277–87
Shipp MA, Ross KN, Tamayo P, Weng AP, Kutok JL, Aguiar RC et al. (2002) Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning. Nat Med 8:68–74
McCormack CJ, Kelly JW, Dorevitch AP (1997) Differences in age and body site distribution of the histological subtypes of basal cell carcinoma. A possible indicator of differing causes. Arch Dermatol 133:593–6
Slamon DJ, Leyland-Jones B, Shak S, Fuchs H, Paton V, Bajamonde A et al. (2001) Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2. N Engl J Med 344:783–92
McMahon AP (2000) More surprises in the Hedgehog signaling pathway. Cell 100:185–8
Sloane JP (1977) The value of typing basal cell carcinomas in predicting recurrence after surgical excision. Br J Dermatol 96:127–32
McMahon AP, Ingham PW, Tabin CJ (2003) Developmental roles and clinical significance of hedgehog signaling. Curr Top Dev Biol 53:1–114
Soehnge H, Ouhtit A, Ananthaswamy ON (1997) Mechanisms of induction of skin cancer by UV radiation. Front Biosci 2:d538–51
Miller JR, Hocking AM, Brown JD, Moon RT (1999) Mechanism and function of signal transduction by the Wnt/beta-catenin and Wnt/Ca2+ pathways. Oncogene 18:7860–72
Sorlie T, Perou CM, Tibshirani R, Aas T, Geisler S, Johnsen H et al. (2001) Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci USA 98:10869–74
Miller S (1991a) Biology of basal cell carcinoma (part I). J Am Acad Dermatol 24:1–13 Miller S (1991b) Biology of basal cell carcinoma (part II). J Am Acad Dermatol 24:161–75 Milovanovic T, Planutis K, Nguyen A, Marsh JL, Lin F, Hope C et al. (2004) Expression of Wnt genes and frizzled 1 and 2 receptors in normal breast epithelium and infiltrating breast carcinoma. Int J Oncol 25:1337–42 Nelson CC, Hoffart D, Gleave ME, Rennie PS (2003) Application of gene microarrays in the study of prostate cancer. Methods Mol Med 81: 299–320 Nilsson M, Unden AB, Krause D, Malmqwist U, Raza K, Zaphiropoulos PG et al. (2000) Induction of basal cell carcinomas and trichoepitheliomas in mice overexpressing GLI-1. Proc Natl Acad Sci USA 97:3438–43 O’Driscoll L, McMorrow J, Doolan P, McKiernan E, Mehta JP, Ryan E et al. (2006) Investigation of the molecular profile of basal cell carcinoma using whole genome microarrays. Mol Cancer 5:74 Orengo IF, Salasche SJ, Fewkes J, Khan J, Thornby J, Rubin F (1997) Correlation of histologic subtypes of primary basal cell carcinoma and number of Moh’s stages required to achieve a tumor-free plane. J Am Acad Dermatol 37:395–7
Strniskova M, Barancik M, Ravingerova T (2002) Mitogen-activated protein kinases and their role in regulation of cellular processes. Gen Physiol Biophys 21:231–55 Tong BJ, Tan J, Tajeda L, Das SK, Chapman JA, DuBois RN et al. (2000) Heightened expression of cyclooxygenase-2 and peroxisome proliferator-activated receptor-delta in human endometrial adenocarcinoma. Neoplasia 2:483–90 Tsuji S, Hosotani R, Yonehara S, Masui T, Tulachan SS, Nakajima S et al. (2003) Endogenous decoy receptor 3 blocks the growth inhibition signals mediated by Fas ligand in human pancreatic adenocarcinoma. Int J Cancer 106:17–25 Tusher VG, Tibshirani R, Chu G (2001) Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci USA 98: 5116–21 Uefuji K, Ichikura T, Mochizuki H (2000) Cyclooxygenase-2 expression is related to prostaglandin biosynthesis and angiogenesis in human gastric cancer. Clin Cancer Res 6:135–8
Pinkus H (1970) The direction of growth of human epidermis. Br J Dermatol 83:556–64
Vadlamudi RK, Wang RA, Talukder AH, Adam L, Johnson R, Kumar R (2000) Evidence of Rab3A expression, regulation of vesicle trafficking, and cellular secretion in response to heregulin in mammary epithelial cells. Mol Cell Biol 20:9092–101
Pomeroy SL, Tamayo P, Gaasenbeek M, Sturla LM, Angelo M, McLaughlin ME et al. (2002) Prediction of central nervous system embryonal tumour outcome based on gene expression. Nature 415:436–42
van’t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M et al. (2002) Gene expression profiling predicts clinical outcome of breast cancer. Nature 415:530–6
Ramaswamy S, Golub TR (2002) DNA microarrays in clinical oncology. J Clin Oncol 20:1932–41
Vincan E, Darcy PK, Smyth MJ, Thompson EW, Thomas RJ, Phillips WA et al. (2005) Frizzled-7 receptor ectodomain expression in a colon cancer cell line induces morphological change and attenuates tumor growth. Differentiation 73:142–53
Reifenberger J, Wolter M, Weber RG, Megahed M, Ruzicka T, Lichter P et al. (1998) Missense mutations in SMOH in sporadic basal cell carcinomas of the skin and primitive neuroectodermal tumors of the central nervous system. Cancer Res 58:1798–803 Riobo NA, Lu K, Ai X, Haines GM, Emerson CP Jr (2006) Phosphoinositide 3-kinase and Akt are essential for Sonic Hedgehog signaling. Proc Natl Acad Sci USA 103:4505–10
Wong CS, Strange RC, Lear JT (2003) Basal cell carcinoma. BMJ 327: 794–8 Wong DJ, Chang HY (2005) Learning more from microarrays: insights from modules and networks. J Invest Dermatol 125:175–82
Rippey JJ (1998) Why classify basal cell carcinomas? Histopathology 32: 393–8
Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J et al. (2002) Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation. Nucleic Acids Res 30:e15
Rosenwald A, Wright G, Chan WC, Connors JM, Campo E, Fisher RI et al. (2002) The use of molecular profiling to predict survival after
Zhivotovsky B, Orrenius S (2006) Carcinogenesis and apoptosis: paradigms and paradoxes. Carcinogenesis 27:1939–45
www.jidonline.org 1805
